Monitoring of the performance of an autoencoder for channel state information compression

Abstract

As an aspect, there is provided an apparatus caused, for example, to: obtain at least one first initial value for monitoring performance of data compression and determine at least one monitoring information, wherein the at least one monitoring information is determined by generating at least one vector representative of a plurality of channel state information estimates, by obtaining at least one pseudo-random index value for selecting at least one element of the at least one vector, or obtaining at least one pseudo-random weight vector for determining at least one vector product with the at least one vector, wherein the at least one pseudo-random index value or the at least one pseudo-random weight vector are obtained by using the at least one first initial value as an input to a pseudo-random sequence generator.

Description

[0001]CHANNEL^STATE^INFORMATION^COMPRESSION TECHNICAL FIELD Various example embodiments relate generally to communication.BACKGROUND The last several generations of mobility have seen spectral efficiency improve substantially through the application of more sophisticated multiple-in- put multiple output (MIMO) techniques. MIMO techniques have been a key technical component for 5G and 5GAdvanced. To support C-band transceiver with more digital RF chains and millime-tre wave transceiver with large number of antenna elements, massive MIMO is oneof the enabling technologies. Beamforming and beam management technologies,including high-precision channel state information (CSI) acquisition for MIMOtransmission, beam training and tracking, beam failure recovery are also intro-duced. Further MIMO enhancements are seen to take place in 6G.BRIEF DESCRIPTION According to some aspects, there is provided the subject matter of theindependent claims. Some further aspects are defined in the dependent claims. Theembodiments that do not fall under the scope of the claims are to be interpreted asexamples useful for understanding the disclosure.LIST OF THE DRAWINGSIn the following, the invention will be described in greater detail with reference to the embodiments and the accompanying drawings, in which Figure 1 presents an example of a network to which one or more em-bodiments are applicable; Figure 2 shows an example of a method; Figure 3 shows another example of a method ;Figure 4 illustrates a signalling chart;Figure 5 illustrates an example of an apparatus;Figure 6 illustrates another example of an apparatus. DESCRIPTION OF EMBODIMENTS The following embodiments are exemplary. Although the specification may refer to “an”, “one”, or “some” embodiment(s) in several locations of the text, this does not necessarily mean that each reference is made to the same embodi- ment(s), or that a particular feature only applies to a single embodiment. Single features of different embodiments may also be combined to provide other embod- iments. For the purposes of the present disclosure, the phrases “at least one of A orB”, “at least one of A and B”, “A and / or B” means (A), (B), or (A and B). For thepurposes of the present disclosure, the phrases “A or B” and “A and / or B” means (A), (B), or (A and B). For the purposes of the present disclosure, the phrase “A, B, and / or C” means (A), (B), (C), (A and B), (A and C), (B and C), or (A, B, and C). It shall be understood that although the terms “first” and “second” etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and sim- ilarly, a second element could be termed a first element, without departing fromthe scope of example embodiments.Embodiments described may be implemented in a radio system, such as one comprising at least one of the following radio access technologies (RATs): Worldwide Interoperability for Micro-wave Access (WiMAX), Global System for Mobile communications (GSM, 2G), GSM EDGE radio access Network (GERAN), General Packet Radio Service (GRPS), Universal Mobile Telecommunication System (UMTS, 3G) based on basic wideband-code division multiple access (W-CDMA), high-speed packet access (HSPA), Long Term Evolution (LTE), LTE-Advanced, and enhanced LTE (eLTE). Term ‘eLTE’ here denotes the LTE evolution that connects to a 5G core. LTE is also known as evolved UMTS terrestrial radio access (EUTRA) or as evolved UMTS terrestrial radio access network (EUTRAN). A term “resource” may refer to radio resources, such as a physical resource block (PRB), a radio frame, a subframe, a time slot, a subband, a frequency region, a sub-carrier, a beam, etc. The term “transmission” and / or “reception” may refer to wirelessly transmit- ting and / or receiving via a wireless propagation channel on radio resources The embodiments are not, however, restricted to the systems / RATs given as an example but a person skilled in the art may apply the solution to othercommunication systems / networks provided with necessary properties. Some ex-amples of a suitable communication networks include a 5G network and / or a 6Gnetwork. The 3GPP solution to 5G is referred to as New Radio (NR).6G is envisagedto be a further development of 5G. NR has been envisaged to use multiple-input-multiple-output (MIMO) multi-antenna transmission techniques, more base sta-tions or nodes than the current network deployments of LTE (a so-called small cell concept), including macro sites operating in co-operation with smaller local area access nodes and perhaps also employing a variety of radio technologies for better coverage and enhanced data rates. 5G will likely be comprised of more than oneradio access technology / radio access network (RAT / RAN), each optimized forcertain use cases and / or spectrum.5G mobile communications may have a wider range of use cases and related applications including video streaming, augmented reality, different ways of data sharing and various forms of machine type applica- tions, including vehicular safety, different sensors and real-time control.5G is ex- pected to have multiple radio interfaces, namely below 6GHz, cmWave and mmWave, and being integrable with existing legacy radio access technologies, such as the LTE. The current architecture in LTE networks is distributed in the radio and centralized in the core network. The low latency applications and services in 5Gmay require bringing the content close to the radio which leads to local break outand multi-access edge computing (MEC).5G enables analytics and knowledge gen- eration to occur at the source of the data. This approach requires leveraging re- sources that may not be continuously connected to a network such as laptops, smartphones, tablets and sensors. MEC provides a distributed computing environ- ment for application and service hosting. It also has the ability to store and process content in close proximity to cellular subscribers for faster response time. Edge computing covers a wide range of technologies such as wireless sensor networks, mobile data acquisition, mobile signature analysis, cooperative distributed peer- to-peer ad hoc networking and processing also classifiable as local cloud / fog com- puting and grid / mesh computing, dew computing, mobile edge computing, cloud- let, distributed data storage and retrieval, autonomic self-healing networks, remote cloud services, augmented and virtual reality, data caching, Internet of Things (massive connectivity and / or latency critical), critical communications (autono- mous vehicles, traffic safety, real-time analytics, time-critical control, healthcare applications). Edge cloud may be brought into RAN by utilizing network function virtualization (NVF) and software defined networking (SDN). Using edge cloud may mean access node operations to be carried out, at least partly, in a server, host or node operationally coupled to a remote radio head or base station comprising radio parts. Network slicing allows multiple virtual networks to be created on top of a common shared physical infrastructure. The virtual networks are then custom- ised to meet the specific needs of applications, services, devices, customers or op- erators. In radio communications, node operations may in be carried out, at least partly, in a central / centralized unit, CU, (e.g. server, host or node) operationally coupled to distributed unit, DU, (e.g. a radio head / node). It is also possible that node operations will be distributed among a plurality of servers, nodes or hosts. Itshould also be understood that the distribution of work between core network op-erations and base station operations may vary depending on implementation. Thus, 5G networks architecture may be based on a so-called CU-DU split. One gNB- CU controls several gNB-DUs. The term ‘gNB’ may correspond in 5G to the eNB in LTE. The gNBs (one or more) may communicate with one or more UEs. The gNB- CU (central node) may control a plurality of spatially separated gNB-DUs, acting at least as transmit / receive (Tx / Rx) nodes. In some embodiments, however, the gNB- DUs (also called DU) may comprise e.g. a radio link control (RLC), medium access control (MAC) layer and a physical (PHY) layer, whereas the gNB-CU (also called a CU) may comprise the layers above RLC layer, such as a packet data convergence protocol (PDCP) layer, a radio resource control (RRC) and an internet protocol (IP) layers. Other functional splits are possible too. It is considered that skilled person is familiar with the OSI model and the functionalities within each layer. In an embodiment, the server or CU may generate a virtual network through which the server communicates with the radio node. In general, virtual networking may involve a process of combining hardware and software network resources and network functionality into a single, software-based administrative entity, a virtual network. Such virtual network may provide flexible distribution of operations between the server and the radio head / node. In practice, any digital signal processing task may be performed in either the CU or the DU and the bound- ary where the responsibility is shifted between the CU and the DU may be selected according to implementation. Some other possible technology advancements to be used are Software- Defined Networking (SDN), Big Data, and all-IP, to mention only a few non-limiting examples. For example, network slicing may be a form of virtual network architec- ture using the same principles behind software defined networking (SDN) and net- work functions virtualisation (NFV) in fixed networks. SDN and NFV may deliver greater network flexibility by allowing traditional network architectures to be par- titioned into virtual elements that can be linked (also through software). Network slicing allows multiple virtual networks to be created on top of a common shared physical infrastructure. The virtual networks are then customised to meet the spe- cific needs of applications, services, devices, customers or operators. The plurality of gNBs (access points / nodes), each comprising the CU and one or more DUs, may be connected to each other via the Xn interface over which the gNBs may negotiate. The gNBs may also be connected over next genera-tion (NG) interfaces to a 5G core network (5GC), which may be a 5G equivalent forthe core network of LTE. Such 5G CU-DU split architecture may be implemented using cloud / server so that the CU having higher layers locates in the cloud and the DU is closer to or comprises actual radio and antenna unit. There are similar plans ongoing for LTE / LTE-A / eLTE as well. When both eLTE and 5G will use similar ar- chitecture in a same cloud hardware (HW), the next step may be to combine soft- ware (SW) so that one common SW controls both radio access networks / technol- ogies (RAN / RAT). This may allow then new ways to control radio resources of both RANs. Furthermore, it may be possible to have configurations where the full pro- tocol stack is controlled by the same HW and handled by the same radio unit as the CU. It should also be understood that the distribution of labour betweencore network operations and base station operations may differ from that of the LTE or even be non-existent. Some other technology advancements probably to be used are Big Data and all-IP, which may change the way networks are being con- structed and managed.5G (or new radio, NR) networks are being designed to sup- port multiple hierarchies, where MEC servers can be placed between the core and the base station or nodeB (gNB). It should be appreciated that MEC can be applied in 4G networks as well. 5G may also utilize satellite communication to enhance or complement the coverage of 5G service, for example by providing backhauling. Possible use cases are providing service continuity for machine-to-machine (M2M) or Internet of Things (IoT) devices or for passengers on board of vehicles, or ensuring service availability for critical communications, and future rail-way / maritime / aeronauti- cal communications. Satellite communication may utilize geostationary earth orbit (GEO) satellite systems, but also low earth orbit (LEO) satellite systems, in partic- ular mega-constellations (systems in which hundreds of (nano)satellites are de- ployed). Each satellite in the mega-constellation may cover several satellite-ena- bled network entities that create on-ground cells. The on-ground cells may be cre- ated through an on-ground relay node or by a gNB located on-ground or in a satel- lite. Embodiments may be also applicable to narrow-band (NB) Internet-of- things (IoT) systems which may enable a wide range of devices and services to be connected using cellular telecommunications bands. NB-IoT is a narrowband radio technology designed for the Internet of Things (IoT) and is one of technologies standardized by the 3rd Generation Partnership Project (3GPP). Other 3GPP IoT technologies also suitable to implement the embodiments include machine type communication (MTC) and eMTC (enhanced Machine-Type Communication). NB- IoT focuses specifically on low cost, long battery life, and enabling a large number of connected devices. The NB-IoT technology is deployed “in-band” in spectrum al-located to Long Term Evolution (LTE) - using resource blocks within a normal LTEcarrier, or in the unused resource blocks within a LTE carrier’s guard-band - or“standalone” for deployments in dedicated spectrum. Embodiments may be also applicable to device-to-device (D2D), ma- chine-to-machine, peer-to-peer (P2P) communications. The embodiments may be also applicable to vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), infra- structure-to-vehicle (I2V), or in general to V2X or X2V communications. Figure 1 illustrates an example of a communication system to which em- bodiments of the invention may be applied. The system may comprise a control node 110 providing one or more cells, such as cell 100, and a control node 112 providing one or more other cells, such as cell 102. Each cell may be, e.g., a macro cell, a micro cell, femto, or a pico cell, for example. In another point of view, the cell may define a coverage area or a service area of the corresponding access node. The control node 110, 112 may be an evolved Node B (eNB) as in the LTE and LTE-A, ng-eNB as in eLTE, gNB of 5G, or any other apparatus capable of controlling radio communication and managing radio resources within a cell. The control node 110, 112 may be called a base station, network node, or an access node. The system may be a cellular communication system composed of a ra- dio access network of access nodes, each controlling a respective cell or cells. The access node 110 may provide user equipment (UE) 120 (one or more UEs) with wireless access to other networks such as the Internet. The wireless access may comprise downlink (DL) communication from the control node to the UE 120 and uplink (UL) communication from the UE 120 to the control node. Additionally, although not shown, one or more local area access nodes may be arranged such that a cell provided by the local area access node at least partially overlaps the cell of the access node 110 and / or 112. The local area access node may provide wireless access within a sub-cell. Examples of the sub-cell may include a micro, pico and / or femto cell. Typically, the sub-cell provides a hot spot within a macro cell. The operation of the local area access node may be controlled by an access node under whose control area the sub-cell is provided. In general, the control node for the small cell may be likewise called a base station, network node, or an access node. There may be a plurality of UEs 120, 122 in the system. Each of them may be served by the same or by different control nodes 110, 112. The UEs 120, 122 may communicate with each other, in case D2D communication interface is established between them. Terms “user device”, “terminal device” or “UE” refer to any end device that may be capable of wireless communication. By way of example rather than limitation, a terminal device may also be referred to as a communication device, user equipment (UE), a Subscriber Station (SS), a Portable Subscriber Station, a Mobile Station (MS), or an Access Terminal (AT). The terminal device may include, but not limited to, a mobile phone, a cellular phone, a smart phone, voice over IP (VoIP) phones, wireless local loop phones, a tablet, a wearable terminal device, a personal digital assistant (PDA), portable computers, desktop computer, image capture terminal devices such as digital cameras, gaming terminal devices, music storage and playback appliances, vehicle-mounted wireless terminal devices, wire- less endpoints, mobile stations, laptop-embedded equipment (LEE), laptop- mounted equipment (LME), USB dongles, smart devices, wireless customer-prem- ises equipment (CPE), an Internet of Things (loT) device, a watch or other weara- ble, a head-mounted display (HMD), a vehicle, a drone, a medical device and appli- cations (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated pro- cessing chain contexts), a consumer electronics device, a device operating on com- mercial and / or industrial wireless networks, and the like. In the following descrip- tion, the terms “terminal device”, “communication device”, “terminal”, “user equip- ment” and “UE” may be used interchangeably. In the case of multiple access nodes in the communication network, the access nodes may be connected to each other with an interface. LTE specifications call such an interface as X2 interface. For IEEE 802.11 network (i.e. wireless local area network, WLAN, WiFi), a similar interface may be provided between access points. An interface between an LTE access point and a 5G access point, or between two 5G access points may be called Xn. Other communication methods between theaccess nodes may also be possible. The access nodes 110 and 112 may be furtherconnected via another interface to a core network 116 of the cellular communica- tion system. The LTE specifications specify the core network as an evolved packet core (EPC), and the core network may comprise a mobility management entity (MME) and a gateway node. The MME may handle mobility of terminal devices in a tracking area encompassing a plurality of cells and handle signalling connectionsbetween the terminal devices and the core network. The gateway node may handledata routing in the core network and to / from the terminal devices. The 5G specifi- cations specify the core network as a 5G core (5GC), and there the core networkmay comprise e.g. an access and mobility management function (AMF) and a userplane function / gateway (UPF), to mention only a few. The AMF may handle termi- nation of non-access stratum (NAS) signalling, NAS ciphering & integrity protec- tion, registration management, connection management, mobility management, access authentication and authorization, security context management. The UPF node may support packet routing & forwarding, packet inspection and QoS han- dling, for example. For downlink communication in MIMO systems, channel state infor- mation (CSI) is needed to effectively leverage the throughput and coverage gains available from the multiple antennas. In current standards, the network node sendspilot signals which are used by the mobile device (UE) to estimate the CSI.CSI estimates are encoded into a compressed form with reducedamount of data and sent upstream to the network node, which decodes the CSI es-timates. In information theory, data compression, source coding, or bit-rate reduc-tion is the process of encoding information using fewer bits than the original rep- resentation of data. Typically, a device that performs data compression is referred to as an encoder, and one that performs the reversal of the process, decompression, as a decoder. Various methods have been proposed and adopted by 3GPP for encod-ing or compressing, transmitting, and decoding or decompressing CSI information.Some examples include the following:• type I codebooks: the CSI is encoded by sending the index of the beam(from a grid of beams) most aligned with the CSI •type II codebooks: the CSI is represented as a linear combination oftwo or more beams •enhanced type II codebook: the CSI is represented as a linear combi-nation of two or more elements of an angle-delay grid A large number of antennas on a network node, gNB, requires a sub- stantial amount of downlink resources for transmitting reference signal for CSI measurements and a significant uplink bandwidth for reporting CSI. CSI feedback enhancement to reduce overhead, improve accuracy, and enhance prediction capa- bilities are under development. CSI compression is one of the technologies under study. The aim is to maximize the compression of CSI feedback at the UE side and minimize information loss during recovery at the gNB side. Thus, as an addition or as an alternative, recent academic research andstudy in 3GPP are considering using trained neural network based functions, such as an autoencoder neural network. An autoencoder artificial intelligence / machine-learning (AI / ML) modelmay achieve both overhead reduction and improved accuracy through joint oper-ation for CSI compression. The autoencoder comprises a matched pair of encoder on the UE and decoder on the gNB, ensuring the decoder can understand the com- pressed output from the UE. In all of these methods, the quality of decoded CSI estimates - how wellit matches the CSI estimates obtained by UE(s) – is not known. However, for exam-ple, knowing the quality of the decoded CSI information may be used by the net-work node to •choose the correct modulation and coding scheme to use for transmis-sion •optimize downlink beamforming weights• change parameters of the feedback scheme. For example, allocatingadditional overhead bits in order to improve the quality of the CSI feedback, as is possible with enhanced type II feedback. Figure 2 depicts an example method. The method may be computer-im-plemented. The method may be carried out by a network node, such as gNB, server,host or any suitable apparatus. The method is suitable for estimating the quality ofdecoded CSI estimates. As simplified, a small number of monitoring bits is trans-mitted upstream, possibly with an encoded CSI feedback message, or indicated theassociated CSI feedback. These bits are processed at the network node, together with the decoded CSI, in order to obtain an estimate of the average quality of the decoded CSI, or performance of the compression of CSI over a past interval of time that may be one or more channel status indication cycles. In block 200, at least one second initial value for monitoring perfor-mance of data compression is obtained.An initial value may be a seed value shared between UE and the networknode enabling a common pseudo-random bit sequence to be available on encoderand decoder ends. A seed is a number that initializes the selection of a sequence bya pseudo-random number generator. A pseudo-random number generator(PRNG), also known as a deterministic random bit generator (DRBG), is an algo- rithm for generating a sequence of numbers whose properties approximate the properties of sequences of random numbers. If given the same seed, a randomnumber generator will generate the same series of random numbers each time theprocess is executed. This feature produces reproducibility.A seed value may be an integer value chosen randomly. It may also be standardized in which case no signalling of the seed value is needed. If signalling isneeded, the seed value may be signalled as a part of CSI configuration or as a partof any control signalling, or in dedicated signalling. The signalling may also com-prise information on quantization for transmitting monitoring information, num- ber of bits in of the monitoring information, information on the duration of the one channel status indication cycle, and / or periodicity of the one or more channel sta- tus indication cycle. As an example, a way to implement pseudo-random sequences isthrough linear-feedback shift registers. In this example, the network node and UEboth have the same type of linear-feedback shift register (LFSR). A network nodemay choose the seed value, such as 001, and transmit that to the UE. Both endsinitialize their LFSR with 001, and both LSFRs will generate the same sequence of pseudo-random bits. In block 202, at least one first monitoring information is received fromone or more user devices, UEs. The at least one first monitoring information is de- termined by generating at least one first vector representative of a plurality of chan-nel state information estimates, by obtaining at least one first pseudo-random in-dex value for selecting at least one element of the at least one first vector, or ob-taining at least one first pseudo-random weight vector for determining at least onevector product with the at least one first vector, wherein the at least one firstpseudo-random index value or the at least one first pseudo-random weight vectorare obtained by using at least one first initial value as an input to a first pseudo- random sequence generator, and wherein the at least one first initial value and the at least one second initial value are at least substantially same. In the following examples, channel estimation and CSI encoding are con-ventional operations that produce an M×N CSI matrix C. An ^ × ^ CSI matrix ^represents channel gains from ^ antenna ports on ^ frequency sub bands. N vec-tors ciof length M are the columns of matrix C.In the first option, index selection, at UE, the bit generator produces abit vector b1 typically in every CSI feedback cycle. The bit vector b1 is used to selecta number I between 1 and NM2. The selected index I and CSI matrix C are used togenerate a monitoring information ^ 1. The index points to the I-th element of vec-tor, vithat is the monitoring information. The encoded CSI matrix q(C) and encodedmonitoring information q(^) are transmitted to the network node.In the second option, weight vector generation, at UE, a weight vector wof length NM2 is generated by using a pseudo-random sequence generator. A vec-tor product is generated of ^ and ^ to define a monitoring information. The vectorproduct is defined as 〈^, ^〉 = ^^^ = ∑^ ^^^ ^^^^ if ^ is a real vector, or 〈^, ^〉 =^^^ = ∑^ ^^^ ^^^^^^^ if ^ is a complex vector, wherein K = NM2. This step could bereferred to as a “weighted sum” of the elements of ^. It is also possible to define more than one monitoring information^, give them (random) weights and generate a sum of these values for defining amonitoring information. The monitoring information may then be a randomizedsummary of ^. The process may be called a random projection. In block 204, at least one second monitoring information is determined(by the network node), wherein the at least one second monitoring information is determined by generating at least one second vector representative of the plurality ofchannel state information estimates as decompressed, by obtaining at least onesecond pseudo-random index value for selecting at least one element of the at leastone second vector, or obtaining at least one second pseudo-random weight vectorfor determining at least one vector product with the at least one second vector,wherein the at least one second pseudo-random index value or the at least one sec- ond pseudo-random weight vector are obtained by using the at least one second initial value as an input to a second pseudo-random sequence generator. The process is at least substantially the same as in the UE for producingmonitoring information suitable for resemblance evaluation. For this purpose, theat least one first initial value and the at least one second initial value are at leastsubstantially same as put forward above. The first pseudo-random sequence gen-erator and the second pseudo-random sequence generator operate in a fashionthat produces at least substantially the same sequence of bits, when seed values,herein called first initial value / second initial value, are at least substantially the same. Based on the method or algorithm used in implementing the pseudo-randomsequence generator, some variance may exist but as long as a selected resemblance evaluation method is applicable, some variation is acceptable. The network node may receive CSI estimates along with receiving the at least one monitoring information. Another option is that UE carries out a con- ventional CSI reporting and transmits the at least one monitoring information indedicated signalling. It is possible to carry out the performance evaluation in everyCSI cycle or less frequently, possibly by indicating the evaluation cycle in a CSI con- figuration message, or triggering the evaluation by a dedicated signalling. In the following examples, channel estimation and CSI encoding and de-coding are conventional operations that produce an M×N CSI matrix ^^ . In general,an ^ × ^ CSI matrix ^ (or estimate ^^ )represents channel gains from ^ antennaports on ^ frequency sub bands. N vectors ci of length M are the columns of matrixC. In the first option, index selection, at the network node, the bit genera-tor produces a bit vector bItypically in every CSI feedback cycle. The bit vector b1is used to select a number I between 1 and NM2. The selected index I and CSI matrixC are used to generate a monitoring information ^I. The index points to the I-thelement of vector, ^I , that is the monitoring information in this example. The en-coded CSI matrix q(C) and encoded monitoring information q(^) are received bythe network node either in a same signalling message or in separate signalling mes-sages. In the second option, weight vector generation, at the network node, aweight vector w of length NM2is generated by using a pseudo-random sequencegenerator. A vector product is generated of ^ and ^ to define a monitoring infor-mation. The vector product is defined as 〈^, ^〉 = ^^^ = ∑^ ^^^ ^^^^ if ^ is a realvector, or 〈^, ^〉 = ^^^ = ∑^ ^^^ ^^^^^^^ if ^ is a complex vector, wherein K = NM2.This step could be referred to as a “weighted sum” of the elements of ^. The weightvector w may be chosen to be sparse to reduce computational complexity of com-puting the vector product. It is possible to define more than one pieces of monitoring information^, give them (random) weights and generate a sum of these values for defining amonitoring information. The monitoring information may then be a randomizedsummary of ^. The process may be called a random projection. As an option, the at least one first monitoring information is received as compressed by a dithered quantization process using the at least one initial value in generating at least one first dithering value by the first pseudo-random sequence generator. In this case, the decoding the at least one first monitoring information further comprises: carrying out a dithered dequantization process using the at least one second initial value in generating at least one second dithering value bythe second pseudo-random sequence generator. The amount of upstream through-put consumed depends on the way ^^is quantized. By using dithered quantization,the accuracy of the monitoring system may be kept acceptable, even only a few bitsof feedback per CSI measurement are used.The monitoring information as well as the CSI report are usually quan-tized for transmission. Quantization causes error. Dithering randomizes this noise. Dither is added before quantization or re-quantization process, in order to de-cor- relate the quantization noise from the input signal and to prevent non-linear be- havior (distortion). Quantization with lesser bit depth requires higher amounts of dither. There are a plurality of different dithering methods, such as Rectangularprobability density function (RPDF) and triangular probability density function(TPDF). Dithered quantization and dequantization require the use of commonpseudo-random bit sequence, b2, at the UE and at the network node. The at leastone first initial value and the at least one second initial value may be used as a seedvalue, respectively. Another option is to use a different seed value for ditheringthan for index selection or weight vector generation. At the encoder, UE side: 1. Use random bits ^^ to generate complex random variable ^(^^).2. Quantize the product ^^ ^(^^) using conventional fixed complex quantiza-tion codebook to obtain ^(^^). At the decoder, network node side: 1. Dequantize ^(^^) using conventional complex quantization codebook to ob-tain ^, an approximation of ^^^(^^). 2. Use random bits ^^ to generate complex random variable ^(^^^).3. Form the output as ^^= .The purpose of the dithering is to make the quan- tization error independent of the input or approximately so. As an example: Quantization: In this scheme, ^^bits are used to represent quantized phase, and ^^bits are used to represent the amplitude. The phase is uniformly quantized, withcode values at ^^ = 2^^ 2^^^ for ^ = 0, … , 2^^ − 1. The amplitude in dB is uni-formly quantized between -10 and 10, with code values at ^^ = −10 +(2^ + 1)102^^^ for ^ = 0, … , 2^^ − 1. For quantization of a complex number ^ =^^^^ , the phase ^ is quantized to the nearest code value ^^ and the amplitude indB, 10 log^^ ^ is quantized to the nearest code value ^^ .In the dequantization step, the indices for phase and amplitude are useto select a pair of values (^^ , ^^), and the complex number is ^ = 10^^d ^^^^^^. Dithering variable: For the dithering variable: the random bits ^^are used to select a ran-dom phase ^ in the interval [0,2^), and a random amplitude ^ (in dB) i^n the range[−10 ⋅ 2^^^ , 10 ⋅ 2^^^ ). The variable is then constructed as In block 206, at least one metric indicative to the performance of the data compression is estimated using a resemblance evaluation of the received at least one first monitoring information and the determined at least one second mon- itoring information. The resemblance evaluation may be carried out by using squared gen-eralized cosine similarity. Asquared generalized cosine similarity (SGCS) may be used as one KPI,or metric, for the performance of the compression. For example, it may be used asintermediate evaluation on AI / ML model performance for CSI. The SGCS calcula-tion for an ^ × ^ CSI matrix ^ and its decoded / decompressed version ^^ may beexpressed as a scalar product of two vectors ^ and ^ of length ^^^, where the vec-tor ^ is a function of ^ and the vector ^ is a function of ^^ . If upstream overhead andcomputation were not important, the mobile device could calculate ^, and send itupstream. The network node could calculate ^^ from ^^ , and then easily calculate theSGCS as Γ^^, ^^^ = ^^^^.As explained above, to save computation resources and bandwidth, auser device may select a random element of ^, ^^ , where ^ is a randomly chosenindex. Then the network node may determine the corresponding element of ^ andcomputes which is an unbiased estimate of the SGCS. For enhanced accu-racy, SGCS estimates may be averaged over a sufficient period of time. SGCS may be defined as where the ^ vectors ^^ of length ^ are the columns of matrix ^, andwhere ^^is the ^-th column of ^. Define ^^ = ^^ / ‖^^‖, then Note that ^, ^, and ^ can be obtained from ^ as ^ = (^ − 1) mod ^ + 1, see from the above for-mula for Γ^^, ^^^ that if the index ^ is chosen uniformly at random from 1 to ^^^,then ^[^^ ^^^] = Γ^^, ^^^. This is why Γ^^ = ^^ ^^^ is an unbiased estimate of Also, since the expected imaginary part of ^^ ^^^ is zero, the estimate Γ^^ = Re(^^ ^^^)is also unbiased. As put forward above, index I is chosen using a pseudo-random se-quence generator. Another example method suitable for the resemblance evaluation is es-timating the distance between the at least one first monitoring information and theat least one second monitoring information. Defining Euclidean distance is one op-tion. For example, suppose that the goal is to evaluate the Euclidian distance be-tween two vectors, ^ and ^, both of which have unit norm ‖^‖^ = ‖^‖^ = 1. Thenthe resemblance evaluation ‖^ − ^‖^ can be estimated as follows. The method de-scribed above for estimating ^^ ^^ can be used to estimate ^^^, and then the resem-blance evaluation can be computed from the formula ‖^ − ^‖^ = 2 − 2 Re(^^^).The method above is suitable for producing adequately accurate esti-mate of the performance of the CSI compression, while saving in signaling and dataprocessing resources. In blocks 208 and 210, in the case the at least one metric indicates aneed for performance adaptation, at least one downlink channel transmission pa-rameter is adjusted, at least one channel state information reporting parameter isadjusted and / or retraining of a machine-learning-based autoencoder used for thedata compression is triggered.The decision to carry out measures to adapt or improve the performance may be based on using a threshold value that may be determinedbased on simulation trials for different use cases. For example, when SGCS is usedas a metric, the values range between 0 and 1, and, depending on the importance of accuracy of CSI values, the threshold may be set closer to or farther away from 1. Some examples of downlink channel transmission parameters com- prise a modulation scheme, a code rate, a number of transmission layers, and / or multiple-input and multiple-output precoding. Some examples of channel state information reporting parameter com- prises a compression level, a quantization level, a limit on the number of com- pressed bits per a signalling message, and / or an indication of a compression method. Figure 3 depicts another example method. The method may be com-puter-implemented. The method may be carried out by a UE, user device, or anysuitable apparatus. The method is suitable for estimating the quality of decoded CSI estimates. As simplified, a small number of monitoring bits is transmitted up-stream, possibly with an encoded CSI feedback message or indicating the associ-ated CSI feedback. These bits are processed at the network node, together with the decoded CSI, in order to obtain an estimate of the average quality of the decoded CSI, or performance of the compression of CSI over a past interval of time that may be one or more channel status indication cycles. In block 300, at least one first initial value for monitoring performanceof data compression is obtained. The at least one first initial value is obtained in amanner providing at least substantial similarity with at least one second initial value used by a network node in the monitoring performance of the data compres- sion. An initial value may be a seed value shared between UE and the net- work node enabling a common pseudo-random bit sequence to be available on en- coder and decoder ends. A seed is a number that initializes the selection of a se-quence by a pseudo-random number generator. A pseudo-random number gener-ator (PRNG), also known as a deterministic random bit generator (DRBG), is an al- gorithm for generating a sequence of numbers whose properties approximate the properties of sequences of random numbers. If given the same seed, a random number generator will generate the same series of random numbers each time the process is executed. This feature produces reproducibility. A seed value may be an integer value chosen randomly. It may also be standardized in which case no signalling of the seed value is needed. If signalling is needed, the seed value may be signalled as a part of CSI configuration or as a part of any control signalling, or in dedicated signalling. The signalling may also com- prise information on quantization for transmitting monitoring information, num- ber of bits in of the monitoring information, information on the duration of the one channel status indication cycle, and / or periodicity of the one or more channel sta- tus indication cycle. As an example, a way to implement pseudo-random sequences is through linear-feedback shift registers. In this example, the network node and UE both have the same type of linear-feedback shift register (LFSR). A network node may choose the seed value, such as 001, and transmit that to the UE. Both ends initialize their LFSR with 001, and both LSFRs will generate the same sequence of pseudo-random bits. In block 302, at least one (first) monitoring information is determinedby a user device. The at least one monitoring information is determined by generating at least one (first) vector representative of a plurality ofchannel state information estimates, by obtaining at least one pseudo-random in- dex value for selecting at least one element of the at least one vector, or obtaining at least one pseudo-random weight vector for determining at least one vector prod- uct with the at least one vector, wherein the at least one pseudo-random index value or the at least one pseudo-random weight vector are obtained by using at least one first initial value as an input to a pseudo-random sequence generator. In the following examples, channel estimation and CSI encoding are con- ventional operations that produce an M×N CSI matrix C. N vectors ciof length M are the columns of matrix C. In the first option, index selection, the bit generator produces a bit vec- tor b1typically in every CSI feedback cycle. The bit vector b1is used to select anumber I between 1 and NM2. The selected index I and CSI matrix C are used togenerate a monitoring information ^ I The index points to the I-th element of vec-tor, vithat is the monitoring information. The encoded CSI matrix q(C) and encoded monitoring information q(^) are transmitted to the network node. In the second option, weight vector generation, a weight vector w oflength NM2 is generated by using a pseudo-random sequence generator. A vectorproduct is generated of ^ and ^ to define a monitoring information. The vectorproduct is defined as 〈^, ^〉 = ^^^ = ∑^ ^^^ ^^^^ if ^ is a real vector, or 〈^, ^〉 =^^^ = ∑^ ^^^ ^^^^^^^ if ^ is a complex vector, wherein K = NM2. This step could be referred to as a “weighted sum” of the elements of ^. It is also possible to define more than one monitoring information^, give them (random) weights and generate a sum of these values for defining amonitoring information. The monitoring information may then be a randomized summary of ^. The process may be called a random projection. As an option, the at least one first monitoring information is transmitted as compressed by a dithered quantization process using the at least one initial value in generating at least one first dithering value by the first pseudo-random sequence generator. In this case, the decoding the at least one first monitoring in- formation further comprises: carrying out a dithered dequantization process using the at least one second initial value in generating at least one second dithering value by the second pseudo-random sequence generator. The monitoring information as well as the CSI report are usually quan- tized for transmission. Quantization causes error. Dithering randomizes this noise. Dither is added before quantization, in order to de-correlate the quantization noise from the input signal and to prevent non-linear behavior (distortion). Quantization with lesser bit depth requires higher amounts of dither. There are a plurality of different dithering methods, such as Rectangular probability density function (RPDF) and triangular probability density function (TPDF). Dithered quantization and dequantization require the use of common pseudo-random bits, b2,at the UE and network node. At the encoder, UE side: 1. Use random bits ^^to generate complex random variable ^(^^). 2. Quantize the product ^^^(^^) using conventional fixed complex quanti- zation codebook to obtain ^(^^). At the decoder, network node side: 1. Dequantize ^(^^) using conventional complex quantization codebook toobtain ^, an approximation of ^^^(^^). 2. Use random bits ^^to complex random variable ^(^^). 3. Form the output The purpose of the dithering is to make the quantization error inde- pendent of the input or approximately so. As an example: Quantization: In this scheme, ^^bits are used to represent quantized phase, and ^^bits are used to represent the amplitude. The phase is uniformly quantized, withcode values at ^^ = 2^^ 2^^^ for ^ = 0, … , 2^^ − 1. The amplitude in dB is uni-formly quantized between -10 and 10, with code values at ^^ = −10 +(2^ + 1)102^^^ for ^ = 0, … , 2^^ − 1. For quantization of a complex number ^ =^^^^ , the phase ^ is quantized to the nearest code value ^^ and the amplitude indB, 10 log^^ ^ is quantized to the nearest code value ^^ .In the dequantization step, the indices for phase and amplito select a pair of values (^^ , ^^), and the complex number is ^ = 10^t ^ude are used ^^^^^^. Dithering variable: For the dithering variable: the random bits ^^are used to select a ran-dom phase ^ in the interval [0,2^), and a random amplitude ^ (in dB) i^n the range[−10 ⋅ 2^^^ , 10 ⋅ 2^^^ ). The variable is then constructed as In block 304, the at least one monitoring information is transmitted tothe network node for the monitoring performance of the data compression. CSI estimates may be transmitted along with transmitting the at leastone monitoring information. Another option is that UE carries out a conventional CSI reporting and transmits the at least one monitoring information in dedicated signalling. It is possible to carry out the performance evaluation in every CSI cycleor less frequently, possibly by UE receiving indication of the evaluation cycle in aCSI configuration message, or triggering the evaluation by a dedicated signalling bythe network node. Depending on the result of the performance monitoring, the user device may receive control signalling from the network node. Control signalling may comprise indication for adjusting at least one downlink channel transmission parameter, adjusting at least one channel state in- formation reporting parameter and / or triggering retraining of a machine-learning- based autoencoder used for the data compression. Some examples of downlink channel transmission parameters com- prise a modulation scheme, a code rate, a number of transmission layers, and / or multiple-input and multiple-output precoding. Some examples of channel state information reporting parameter com- prises a compression level, a quantization level, a limit on the number of com- pressed bits per a signalling message, and / or an indication of a compression method. Figure 4 depicts examples of signalling in relation to the methods, em-bodiments and examples described above by means of Figures 2 and 3.Network node 110 may transmit to user device, UE 120, CSI reporting configuration, 400, comprising, for example, the frequency of reporting and code- book to use. The network node transmits channel state information reference signal,CSI-RS, in the Downlink for user device to carry out measurements for channel es-timation. The user device reports CSI parameters to the network node as feedback.The CSI feedback includes several parameters, such as channel quality indicator(CQI), precoding matrix indicator (PMI), and rank indicator (RI). Upon receivingthe CSI parameters, the gNB schedules downlink data transmissions (such as mod- ulation scheme, code rate, number of transmission layers, and MIMO precoding) accordingly. The CSI reporting configuration may also comprise monitoring configu-ration such as one or more initial values or seed values for a pseudo-random se-quence generator and / or reporting configuration such as information on quantiza- tion, the number of bits for monitoring information and / or frequency or periodic- ity of the reporting. The configuration may be sent periodically or at least some of the parameters may be configured in advance and be valid for several reporting rounds. Monitoring information may be determined with a longer periodicity than the CSI cycle, or it may be triggered with dedicated signalling. Monitoring configu- ration may be transmitted in a dedicated signalling as well. As an option, control signal 400 is a monitoring request triggering per- formance monitoring. The monitoring request may comprise the monitoring con- figuration. User device, 120, may transmit acknowledgement (ACK) for the report- ing request, 402, when the CSI reporting configuration comprises monitoring con- figuration, or control signal 400 is a monitoring request. In this example, once the network node receives the acknowledgementfor CSI configuration and monitoring configuration, it begins to (periodically)transmit reference signal (CSI-RS) 404 on the physical layer for CSI measurements.The user device prepares 406 a CSI report and transmits it upstream to the net-work node 408. The CSI report may also comprise monitoring information associ-ated with that report, or the monitoring information may be transmitted in a dedi- cated signaling 410. The monitoring information may be determined as described above by means of Figures 2 and 3. Pattern, 404, 406, 408, 410, in different combi-nations, may continue in each feedback cycle, or, as an option, the monitoring in-formation may be transmitted less frequently, until the end of the session or until CSI-RS reporting configuration is changed. Interrupting thedetermination / transmission of the monitoring information is also possible, for ex-ample to save resources, by transmitting corresponding control signal. It is also an option to change the monitoring information configuration by transmitting a mon- itoring (re)configuration message. The monitoring information determination mayalso be triggered by control signaling, such as transmitting a reporting request (see400), on a need basis.The network node estimates performance of the data compression us-ing the monitoring information it receives from the user device (first) and moni-toring information it determines (second) 412. Depending on the result of the per- formance monitoring, the user device may receive control signaling from the net-work node 414. The control signaling may be a configuration signaling that maycomprise indication for adjusting at least one downlink channel transmission pa- rameter, adjusting at least one channel state information reporting parameter and / or triggering retraining of a machine-learning-based autoencoder used for the data compression. An embodiment, as shown in Figure 5, provides an apparatus 50 com- prising a control circuitry (CTRL) 52, such as at least one processor, and at least one memory 54 storing instructions that, when executed by the at least one pro- cessor, cause the apparatus at least to carry out any one of the above-described processes. In an example, the at least one memory and the computer program code (software), are configured, with the at least one processor, to cause the apparatus to carry out any one of the above-described processes. The memory may be imple- mented using any suitable data storage technology, such as semiconductor-based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, fixed memory and removable memory. The memory may comprise a database for storing data. The apparatus may be caused to execute at least some of the function- alities of the above described processes, such as features described by means of Figure 3. For instance, circuitry 52 may comprise circuitry 520 for determining monitoring information and circuitry 522 for estimating CSI. As another option,memory 54 may store (portions of) computer program for carrying out functional-ities described. At least one initial value may be received via radio interface 56,stored and retrieved from memory 54 or determined by circuitry 52, 520. In an embodiment, the apparatus 50 may comprise user device, UE, e.g. a user terminal (UT), a computer (PC), a laptop, a tabloid computer, a cellular phone, a mobile phone, a communicator, a smart phone, a palm computer, a mobile transportation apparatus (such as a car), a household appliance, or any other com- munication apparatus, commonly called as UE in the description. Alternatively, the apparatus is comprised in such a terminal device. Further, the apparatus may be orcomprise a module (to be coupled to the UE) providing connectivity, such as a plug-in unit, an “USB dongle”, or any other kind of a unit. The unit may be installed either inside the UE or attached to the UE with a connector or even wirelessly. In an embodiment, the apparatus 50 is or is comprised in the UE 120.The apparatus may be caused to execute some of the functionalities of the abovedescribed processes, such as features described by means of Figure 3.The apparatus may further comprise a radio interface (transmitter / re-ceiver, TRX) 56 comprising hardware and / or software for realizing communica-tion connectivity according to one or more communication protocols. The TRX mayprovide the apparatus with communication capabilities to access the radio accessnetwork, for example. The apparatus may also comprise a user interface 58 comprising, for example, at least one keypad, a microphone, a touch display, a display, a speaker, etc. The user interface may be used to control the apparatus by the user. An embodiment, as shown in Figure 6, provides an apparatus 60 com-prising a control circuitry (CTRL) 62, such as at least one processor, and at least one memory 64 storing instructions that, when executed by the at least one pro- cessor, cause the apparatus at least to carry out any one of the above-described processes. In an example, the at least one memory and the computer program code (software), are configured, with the at least one processor, to cause the apparatus to carry out any one of the above-described processes. The memory may be imple- mented using any suitable data storage technology, such as semiconductor-based memory devices, flash memory, magnetic memory devices and systems, optical memory devices and systems, fixed memory and removable memory. The memory may comprise a database for storing data. In an embodiment, the apparatus 60 may be or be comprised in a net-work node, such as in gNB / gNB-CU / gNB-DU or edge cloud. In an embodiment, theapparatus is or is comprised in the network node 110. The apparatus may be caused to execute at least some of the functionalities of the above described pro- cesses, such as features described by means of Figure 2. For instance, circuitry 62 may comprise circuitry 620 for determining at least one second monitoring infor-mation and circuitry 622 for estimating at least one metric indicative to the perfor-mance of the data compression and / or adjusting at least one downlink channel transmission parameter. As another option, memory 64 may store (portions of) computer program for carrying out functionalities described. The apparatus may further comprise communication interface (trans- mitter / receiver, TRX) 66 comprising hardware and / or software for realizing com- munication connectivity according to one or more communication protocols. The TRX may provide the apparatus with communication capabilities with at least one user equipment, for example. The apparatus may also comprise a user interface 68 comprising, for example, at least one keypad, a microphone, a touch display, a display, a speaker, etc. The user interface may be used to control the apparatus by the user. In an embodiment, a CU-DU (central unit – distributed unit) architec-ture is applied. In such case the apparatus 60 may be comprised in a central unit(e.g. a control unit, an edge cloud server, a server) operatively coupled (e.g. via a wireless or wired network) to a distributed unit (e.g. a remote radio head / node). That is, the central unit (e.g. an edge cloud server) and the radio node may be stand- alone apparatuses communicating with each other via a radio path or via a wired connection. Alternatively, they may be in a same entity communicating via a wired connection, etc. The edge cloud or edge cloud server may serve a plurality of radionodes or a radio access networks. In an embodiment, at least some of the describedprocesses may be performed by the central unit. In another embodiment, the ap- paratus may be instead comprised in the distributed unit, and at least some of the described processes may be performed by the distributed unit. In an embodiment,the execution of at least some of the functionalities of the apparatus 60 may beshared between two physically separate devices (DU and CU) forming one opera- tional entity. Therefore, the apparatus may be seen to depict the operational entity comprising one or more physically separate devices for executing at least some of the described processes. In an embodiment, the apparatus controls the execution of the processes, regardless of the location of the apparatus and regardless of where the processes / functions are carried out. An apparatus carrying out at least some of the embodiments described may comprise at least one processor and at least one memory including a computer program code, wherein the at least one memory and the computer program code are configured, with the at least one processor, to cause the apparatus to carry out the functionalities according to any one of the embodiments described. According to an aspect, when the at least one processor executes the computer program code, the computer program code causes the apparatus to carry out the functionalities according to any one of the embodiments described. According to another embod- iment, the apparatus carrying out at least some of the embodiments comprises the at least one processor and at least one memory including a computer program code, wherein the at least one processor and the computer program code may carry at least some of the functionalities according to any one of the embodiments de- scribed. Accordingly, the at least one processor, the memory, and the computer program code may form processing means for carrying out at least some of the em- bodiments described. According to yet another embodiment, the apparatus carry- ing out at least some of the embodiments comprises a circuitry including at least one processor and at least one memory including computer program code. When activated, the circuitry causes the apparatus to perform the at least some of thefunctionalities according to any one of the embodiments or examples described.As used in this application, the term ‘circuitry’ refers to all of the follow- ing: (a) hardware-only circuit implementations, such as implementations in only analog and / or digital circuitry, and (b) combinations of circuits and soft-ware (and / or firmware), such as (as applicable): (i) a combination of processor(s) or (ii) portions of processor(s) / software including digital signal processor(s), software, and memory(ies) that work together to cause an apparatus to perform various functions, and (c) circuits, such as a microprocessor(s) or a portion of a micropro- cessor(s), that require software or firmware for operation, even if the software or firmware is not physically present. This definition of ‘circuitry’ applies to all uses of this term in this application. As a further example, as used in this application, the term ‘circuitry’ would also cover an implementation of merely a processor (or mul- tiple processors) or a portion of a processor and its (or their) accompanying soft- ware and / or firmware. The term ‘circuitry’ would also cover, for example and if applicable to the particular element, a baseband integrated circuit or applications processor integrated circuit for a mobile phone or a similar integrated circuit in a server, a cellular network device, or another network device. In an embodiment, at least some of the processes described above by means of Figures 2, 3 and 4 may be carried out by an apparatus comprising corre- sponding means for carrying out at least some of the described processes. Some example means for carrying out the processes may include at least one of the fol-lowing: detector, processor (including dual-core and multiple-core processors),digital signal processor, controller, receiver, transmitter, encoder, decoder,memory, RAM, ROM, software, firmware, display, user interface, display circuitry,user interface circuitry, user interface software, display software, circuit, antenna, antenna circuitry, and circuitry. An example of the apparatus may comprise means (62, 620, 64) for ob- taining, by a network node, at least one second initial value for monitoring perfor- mance of data compression, means (66) for receiving at least one first monitoring information, wherein the at least one first monitoring information is determinedby generating at least one first vector representative of a plurality of channel stateinformation estimates, by obtaining at least one first pseudo-random index value for selecting at least one element of the at least one first vector, or obtaining at least one first pseudo-random weight vector for determining at least one vector product with the at least one first vector, wherein the at least one first pseudo-random in- dex value or the at least one first pseudo-random weight vector are obtained by using at least one first initial value as an input to a first pseudo-random sequence generator, and wherein the at least one first initial value and the at least one second initial value are at least substantially same, means (62, 620) for determining, at least one second monitoring information (by the network node), wherein the atleast one second monitoring information is determined by generating at least onesecond vector representative of the plurality of channel state information esti- mates as decompressed, by obtaining at least one second pseudo-random index value for selecting at least one element of the at least one second vector, or obtain- ing at least one second pseudo-random weight vector for determining at least one vector product with the at least one second vector, wherein the at least one second pseudo-random index value or the at least one second pseudo-random weight vec- tor are obtained by using the at least one second initial value as an input to a second pseudo-random sequence generator, means (62, 622) for estimating at least one metric indicative to the performance of the data compression using a resemblance evaluation of the received at least one first monitoring information and the deter- mined at least one second monitoring information, and, means (62, 622) for adjust- ing, in the case the at least one metric indicates a need for performance adaptation,at least one downlink channel transmission parameter or at least one channel stateinformation reporting parameter, and / or means (62, 622) for triggering retraining of a machine-learning-based autoencoder used for the data compression. Another example of an apparatus comprise means (52, 520, 54, 56) for obtaining at least one first initial value for monitoring performance of data com-pression, wherein the at least one first initial value is obtained in a manner provid-ing at least substantial similarity with at least one second initial value used by a network node in the monitoring performance of the data compression, means (52, 522) for determining at least one first monitoring information by a user device,wherein the at least one monitoring information is determined by generating atleast one vector representative of a plurality of channel state information esti- mates, by obtaining at least one pseudo-random index value for selecting at leastone element of the at least one vector, or obtaining at least one pseudo-randomweight vector for determining at least one vector product with the at least one vec- tor, wherein the at least one pseudo-random index value or the at least one pseudo- random weight vector are obtained by using the at least one first initial value as aninput to a pseudo-random sequence generator, and means (56) for transmitting theat least one monitoring information to the network node for the monitoring per- formance of the data compression. A term non-transitory, as used herein, is a limitation of the medium it- self (i.e. tangible, not a signal) as opposed to a limitation on data storage persis- tency (e.g. RAM vs. ROM). The techniques and methods described herein may be implemented by various means. For example, these techniques may be implemented in hardware (one or more devices), firmware (one or more devices), software (one or more modules), or combinations thereof. For a hardware implementation, the appa- ratus(es) of embodiments may be implemented within one or more application- specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programma- ble gate arrays (FPGAs), processors, controllers, micro-controllers, microproces- sors, other electronic units designed to perform the functions described herein, or a combination thereof. For firmware or software, the implementation can be car- ried out through modules of at least one chip set (e.g. procedures, functions, and so on) that perform the functions described herein. The software codes may be stored in a memory unit and executed by processors. The memory unit may be imple- mented within the processor or externally to the processor. In the latter case, it can be communicatively coupled to the processor via various means, as is known in the art. Additionally, the components of the systems described herein may be rear- ranged and / or complemented by additional components in order to facilitate the achievements of the various aspects, etc., described with regard thereto, and they are not limited to the precise configurations set forth in the given figures, as will be appreciated by one skilled in the art. Embodiments as described may also be carried out in the form of a com- puter process defined by a computer program or portions thereof. Embodiments and examples of the methods described may be carried out by executing at least one portion of a computer program comprising corresponding instructions. The computer program may be in source code form, object code form, or in some inter- mediate form, and it may be stored in some sort of carrier, which may be any entity or device capable of carrying the program. For example, the computer program may be stored on a computer program distribution medium readable by a com- puter or a processor. The computer program medium may be, for example but not limited to, a record medium, computer memory, read-only memory, electrical car- rier signal, telecommunications signal, and software distribution package, for ex- ample. The computer program medium may be a non-transitory medium. Coding of software for carrying out the embodiments as shown and described is well within the scope of a person of ordinary skill in the art. Some aspects comprise: According to a first aspect, there is provided an apparatus, comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: obtain, by a network node, at least one second initial value for monitoring perfor- mance of data compression; receive at least one first monitoring information, wherein the at least one first mon- itoring information is determined by generating at least one first vector representative of a plurality of channel state in- formation estimates by obtaining at least one first pseudo-random index value for selecting at least one element of the at least one first vector, or obtaining at least one first pseudo-random weight vector for deter-mining at least one vector prod- uct with the at least one first vector, wherein the at least one first pseudo-random index value or the at least one first pseudo-random weight vector are obtained by using at least one first initial value as an input to a first pseudo-random sequence generator, and wherein the at least one first initial value and the at least one second initial value are at least substantially same; determine at least one second monitoring information, wherein the at least one second monitoring information is determined by generating at least one second vector representative of the plurality of channel state information estimates as decompressed, by obtaining at least one second pseudo-random index value for selecting at least one element of the at least one second vector, or obtaining at least one second pseudo-random weight vector for determining at least one vector product with the at least one second vector, wherein the at least one second pseudo-random index value or the at least one sec- ond pseudo-random weight vector are obtained by using the at least one second initial value as an input to a second pseudo-random sequence generator, and estimate at least one metric indicative to the performance of the data compression using a resemblance evaluation of the received at least one first monitoring infor- mation and the determined at least one second monitoring in-formation, and in thecase the at least one metric indicates a need for performance adaptation,adjust at least one downlink channel transmission parameter or at least one chan- nel state information reporting parameter, and / or trigger retraining of a machine- learning-based autoencoder used for the data compression.According to a second aspect, there is provided the apparatus of the first aspect,wherein the at least one first monitoring information is received as compressed by a dithered quantization process using the at least one initial value in generating at least one first dithering value by the first pseudo-random sequence generator, and wherein the decoding the at least one first monitoring information further com-prises causing the apparatus to: carry out a dithered dequantization process usingthe at least one second initial value in generating at least one second dithering value by the second pseudo-random sequence generator.According to a third aspect, there is provided the apparatus, of the first or secondaspect, further comprising causing the apparatus to: average the at least one metric estimated during one or more channel status indication cycles.According to a fourth aspect, there is provided the apparatus of any of the preced-ing aspect, wherein the resemblance evaluation is carried out by using squared generalized cosine similarity.According to a fifth aspect, there is provided the apparatus of any preceding aspect,wherein the first pseudo-random sequence generator and the second pseudo-ran- dom sequence generator operate in a fashion that produces at least substantially same sequence when the at least one first initial value and the at least one second initial value are at least substantially same.According to sixth aspect, there is provided the apparatus, of any preceding aspect,wherein the at least one downlink channel transmission parameters comprise a modulation scheme, a code rate, a number of transmission layers, and / or multiple- input and multiple-output precoding.According to a seventh aspect, there is provided the apparatus of any precedingaspect, wherein the at least one channel state information reporting parametercomprises a compression level, a quantization level, a limit on the number of com-pressed bits per a signalling message, and / or an indication of a compression method.According to an eight aspect, there is provided the apparatus of any preceding as-pect, further comprising causing the apparatus to: indicate the at least one first in- itial value to at least one user device as a part of channel state information report- ing configuration message, or using a dedicated signalling message.According to a ninth aspect, there is provided the apparatus of the eight aspect,wherein the message further comprises information on quantization for the at least one first monitoring information, number of bits in the at least one first monitoring information, information on the duration of the one channel status indication cycle, and / or periodicity of the one or more channel status indication cycle. According to a tenth aspect, there is provided an apparatus, comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: obtain, by a user device, at least one first initial value for monitoring performance of data compression, wherein the at least one first initial value is obtained in a man- ner providing at least substantial similarity with at least one second initial value used by a network node in the monitoring performance of the data compression, determine at least one monitoring information, wherein the at least one monitor- ing information is determined by generating at least one vector rep-resentative of a plurality of channel state information estimates, by obtaining at least one pseudo- random index value for selecting at least one element of the at least one vector, or obtaining at least one pseudo-random weight vector for de-termining at least one vector product with the at least one vector, wherein the at least one pseudo- random index value or the at least one pseudo-random weight vector are obtained by using the at least one first initial value as an input to a pseudo-random sequence generator, and transmit the at least one monitoring information to the network node for the mon- itoring performance of the data compression.According to an eleventh aspect, there is provided the apparatus of the tenth as-pect, further comprising causing the apparatus to: carry out, for the transmission, a dithered dequantization process for the at least one monitoring information us- ing the at least one first initial value in generating at least one dithering value by the pseudo-random sequence generator.According to a twelfth aspect, there is provided the apparatus of the tenth or theeleventh aspects, wherein the pseudo-random sequence generator operates in a fashion that produces at least substantially same sequence with the at least one first initial value than a second pseudo-random generator in the network node with the at least one second initial value.According to a thirteenth aspect, there is provided the apparatus of any precedingaspects the tenth to the twelfth, further comprising causing the apparatus to: receive, from the network node, a control message in association with the monitor- ing the performance of data compression, and transmit an acknowledgement to the control message.According to a fourteenth aspect, there is provided the apparatus of the thirteenthaspect, further comprising causing the apparatus to: receive an indication on the at least one first initial value in the control message.According to a fifteenth aspect, there is provided the apparatus of the thirteenthor the fourteenth aspects, wherein the control message is a configuration message or a reporting request message.According to a sixteenth aspect, there is provided a method comprising:obtaining, by a network node, at least one second initial value for monitoring per- formance of data compression; receiving at least one first monitoring information, wherein the at least one first monitoring information is determined by generating at least one first vector representative of a plurality of channel state in- formation estimates by obtaining at least one first pseudo-random index value for selecting at least one element of the at least one first vector, or obtaining at least one first pseudo-random weight vector for deter-mining at least one vector prod- uct with the at least one first vector, wherein the at least one first pseudo-random index value or the at least one first pseudo-random weight vector are obtained by using at least one first initial value as an input to a first pseudo-random sequence generator, and wherein the at least one first initial value and the at least one second initial value are at least substantially same; determining at least one second monitoring information, wherein the at least one second monitoring information is determined by generating at least one second vector representative of the plurality of channel state information estimates as decompressed, by obtaining at least one second pseudo-random index value for selecting at least one element of the at least one second vector, or obtaining at least one second pseudo-random weight vector for determining at least one vector product with the at least one second vector, wherein the at least one second pseudo-random index value or the at least one sec- ond pseudo-random weight vector are obtained by using the at least one second initial value as an input to a second pseudo-random sequence generator, and estimating at least one metric indicative to the performance of the data compres- sion using a resemblance evaluation of the received at least one first monitoring information and the determined at least one second monitoring in-formation, and in the case the at least one metric indicates a need for performance adaptation, adjusting at least one downlink channel transmission parameter or at least one channel state information reporting parameter, and / or trigger retraining of a ma- chine-learning-based autoencoder used for the data compression.According to a seventeenth aspect, there is provided the method of the sixteenthaspect, wherein the at least one first monitoring information is received as com- pressed by a dithered quantization process using the at least one initial value in generating at least one first dithering value by the first pseudo-random sequence generator, and wherein the decoding the at least one first monitoring information further comprises causing the apparatus to: carry out a dithered dequantization process using the at least one second initial value in generating at least one second dithering value by the second pseudo- random sequence generator.According to an eighteenth aspect, there is provided the method of the sixteenthor seventeenth aspects, further comprising: averaging the at least one metric esti- mated during one or more channel status indication cycles.According to a nineteenth aspect, there is provided the method of any of the pre-ceding aspects the sixteenth to the eighteenth, wherein the resemblance evaluationis carried out by using squared generalized cosine similarity.According to a twentieth aspect, there is provided the method of any preceding as-pect the sixteenth to the nineteenth, wherein the first pseudo-random sequence generator and the second pseudo-random sequence generator operate in a fashion that produces at least substantially same sequence when the at least one first ini- tial value and the at least one second initial value are at least substantially same .According to a twenty first aspect, there is provided the method of any precedingaspect the sixteenth to the twentieth, wherein the at least one downlink channel transmission parameters comprise a modulation scheme, a code rate, a number of transmission layers, and / or multiple-input and multiple-output precoding.According to a twenty second aspect, there is provided the method of any preced-ing aspect the sixteenth to the twenty first, wherein the at least one channel state information reporting parameter comprises a compression level, a quantization level, a limit on the number of compressed bits per a signal-ling message, and / or an indication of a compression method.According to a twenty third aspect, there is provided the method of any precedingaspect the sixteenth to the twenty second, further comprising: indicating the atleast one first initial value to at least one user device as a part of channel state in- formation reporting configuration message, or using a dedicated signalling mes- sage.According to a twenty fourth aspect, there is provided the method of the twentythird aspect, wherein the message further comprises information on quantization for the at least one first monitoring information, number of bits in the at least one first monitoring information, information on the duration of the one channel status indication cycle, and / or periodicity of the one or more channel status indication cycle.According to a twenty fifth aspect, there is provided a method, comprising:obtaining, by a user device, at least one first initial value for monitoring perfor- mance of data compression, wherein the at least one first initial value is obtained in a manner providing at least substantial similarity with at least one second initial value used by a network node in the monitoring performance of the data compres- sion, determining at least one monitoring information, wherein the at least one moni- toring information is determined by generating at least one vector representative of a plurality of channel state information estimates, by obtaining at least one pseudo-random index value for selecting at least one element of the at least one vector, or obtaining at least one pseudo-random weight vector for determining at least one vector product with the at least one vector, wherein the at least one pseudo-random index value or the at least one pseudo-random weight vector are obtained by using the at least one first initial value as an input to a pseudo-random sequence generator, and transmitting the at least one monitoring information to the network node for the monitoring performance of the data compression.According to a twenty sixth aspect, there is provided the method of the twenty fifthaspect, further comprising: carrying out, for the transmission, a dithered dequan-tization process for the at least one monitoring information using the at least one first initial value in generating at least one dithering value by the pseudo-random sequence generator.According to a twenty seventh aspect, there is provided the method of the twentyfifth aspect or the twenty sixth aspect, wherein the pseudo-random sequence gen- erator operates in a fashion that produces at least substantially same sequence with the at least one first initial value than a second pseudo-random generator in the network node with the at least one second initial value.According to a twenty eight aspect, there is provided the method of any precedingaspect the twenty fifth to the twenty sixth , further comprising: receiving, from the network node, a control message in association with the monitoring the perfor- mance of data compression, and transmitting an acknowledgement to the control message.According to a twenty nineth aspect, there is provided the method of the twentyeight aspect further comprising: receiving an indication on the at least one first in- itial value in the control message. According to a thirtieth aspect, there is provided the method of the twenty eight or twenty nineth aspect, wherein the control message is a configuration message or a reporting request message. According to a thirty first aspect, there is provided an apparatus, comprising means for performing the aspects according to any of the aspects the sixteenth to the twenty fourth or the twenty fifth to the thirtieth. Even though the invention has been described above with reference to an example according to the accompanying drawings, it is clear that the invention is not restricted thereto but can be modified in several ways within the scope of the appended claims. Therefore, all words and expressions should be interpreted broadly and they are intended to illustrate, not to restrict, the embodiment. It will be obvious to a person skilled in the art that, as technology advances, the inventive concept can be implemented in various ways. Further, it is clear to a person skilled in the art that the described embodiments may, but are not required to, be com- bined with other embodiments in various ways.

Claims

CLAIMS 1. An apparatus, comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: obtain, by a network node, at least one second initial value for monitor- ing performance of data compression; receive at least one first monitoring information, wherein the at leastone first monitoring information is determined by generating at least one first vector representative of a plurality of chan-nel state information estimates by obtaining at least one first pseudo-random index value for selecting at least one element of the at least one first vector, or obtaining at least one first pseudo-random weight vector for determining at least one vector product with the at least one first vector, wherein the at least one first pseudo-ran- dom index value or the at least one first pseudo-random weight vector are obtained by using at least one first initial value as an input to a first pseudo-random se- quence generator, and wherein the at least one first initial value and the at least one second initial value are at least substantially same; determine at least one second monitoring information, wherein the atleast one second monitoring information is determined by generating at least one second vector representative of the plurality of channel state information estimates as decompressed, by obtaining at least one second pseudo-random index value for selecting at least one element of the at leastone second vector, or obtaining at least one second pseudo-random weight vectorfor determining at least one vector product with the at least one second vector, wherein the at least one second pseudo-random index value or the at least one sec- ond pseudo-random weight vector are obtained by using the at least one secondinitial value as an input to a second pseudo-random sequence generator, andestimate at least one metric indicative to the performance of the datacompression using a resemblance evaluation of the received at least one first mon- itoring information and the determined at least one second monitoring infor-mation, and in the case the at least one metric indicates a need for performanceadaptation, adjust at least one downlink channel transmission parameter or at least one channel state information reporting parameter, and / or trigger retraining of amachine-learning-based autoencoder used for the data compression.

2. The apparatus of claim 1, wherein the at least one first monitoring information is received as compressed by a dithered quantization process using the at least one initial value in generating at least one first dithering value by the first pseudo-random sequence generator, and wherein the decoding the at leastone first monitoring information further comprises causing the apparatus to:carry out a dithered dequantization process using the at least one sec- ond initial value in generating at least one second dithering value by the second pseudo-random sequence generator.

3. The apparatus of claim 1 or 2, further comprising causing the appa-ratus to: average the at least one metric estimated during one or more channelstatus indication cycles.

4. The apparatus of any preceding claim, wherein the resemblance eval-uation is carried out by using squared generalized cosine similarity.

5. The apparatus of any preceding claim, wherein the first pseudo-ran-dom sequence generator and the second pseudo-random sequence generator op- erate in a fashion that produces at least substantially same sequence when the at least one first initial value and the at least one second initial value are at least sub- stantially same.

6. The apparatus of any preceding claim, wherein the at least one down-link channel transmission parameters comprise a modulation scheme, a code rate, a number of transmission layers, and / or multiple-input and multiple-output pre- coding.

7. The apparatus of any preceding claim, wherein the at least one chan- nel state information reporting parameter comprises a compression level, a quan- tization level, a limit on the number of compressed bits per a signalling message, and / or an indication of a compression method.

8. The apparatus of any preceding claim, further comprising causing theapparatus to: indicate the at least one first initial value to at least one user device as apart of channel state information reporting configuration message, or using a ded- icated signalling message.

9. The apparatus of claim 8, wherein the message further comprises in-formation on quantization for the at least one first monitoring information, number of bits in the at least one first monitoring information, information on the duration of the one channel status indication cycle, and / or periodicity of the one or more channel status indication cycle.

10. An apparatus, comprising: at least one processor; and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to: obtain, by a user device, at least one first initial value for monitoringperformance of data compression, wherein the at least one first initial value is ob- tained in a manner providing at least substantial similarity with at least one second initial value used by a network node in the monitoring performance of the data compression, determine at least one monitoring information, wherein the at least onemonitoring information is determined by generating at least one vector repre-sentative of a plurality of channel state information estimates, by obtaining at least one pseudo-random index value for selecting at least one element of the at least one vector, or obtaining at least one pseudo-random weight vector for determining at least one vector product with the at least one vector, wherein the at least one pseudo-random index value or the at least one pseudo-random weight vector are obtained by using the at least one first initial value as an input to a pseudo-random sequence generator, and transmit the at least one monitoring information to the network nodefor the monitoring performance of the data compression.

11. The apparatus of claim 10, further comprising causing the apparatus to: carry out, for the transmission, a dithered dequantization process forthe at least one monitoring information using the at least one first initial value ingenerating at least one dithering value by the pseudo-random sequence generator.

12. The apparatus of claim 10 or 11, wherein the pseudo-random se-quence generator operates in a fashion that produces at least substantially samesequence with the at least one first initial value than a second pseudo-random gen-erator in the network node with the at least one second initial value.

13. The apparatus of any preceding claim 10 to 12, further comprising causing the apparatus to: receive, from the network node, a control message in association withthe monitoring the performance of data compression, andtransmit an acknowledgement to the control message.

14. The apparatus of claim 13, further comprising causing the apparatus to: receive an indication on the at least one first initial value in the control message.

15. The apparatus of claim 13 or 14, wherein the control message is aconfiguration message or a reporting request message.

16. A method, comprising: obtaining, by a network node, at least one second initial value for mon- itoring performance of data compression; receiving at least one first monitoring information, wherein the at leastone first monitoring information is determined by generating at least one first vector representative of a plurality of chan- nel state information estimates by obtaining at least one first pseudo-random index value for selecting at least one element of the at least one first vector, or obtaining at least one first pseudo-random weight vector for deter-mining at least one vector product with the at least one first vector, wherein the at least one first pseudo-ran- dom index value or the at least one first pseudo-random weight vector are obtained by using at least one first initial value as an input to a first pseudo-random se- quence generator, and wherein the at least one first initial value and the at least one second initial value are at least substantially same;determining at least one second monitoring information, wherein the atleast one second monitoring information is determined by generating at least one second vector representative of the plurality of channel state information estimates as decompressed, by obtaining at least one second pseudo-random index value for selecting at least one element of the at least one second vector, or obtaining at least one second pseudo-random weight vector for determining at least one vector product with the at least one second vector, wherein the at least one second pseudo-random index value or the at least one sec- ond pseudo-random weight vector are obtained by using the at least one second initial value as an input to a second pseudo-random sequence generator, and estimating at least one metric indicative to the performance of the datacompression using a resemblance evaluation of the received at least one first mon- itoring information and the determined at least one second monitoring in-for-mation, and in the case the at least one metric indicates a need for performanceadaptation, adjusting at least one downlink channel transmission parameter or atleast one channel state information reporting parameter, and / or trigger retraining of a machine-learning-based autoencoder used for the data compression.

17. The method of claim 16, wherein the at least one first monitoringinformation is received as compressed by a dithered quantization process using the at least one initial value in generating at least one first dithering value by the first pseudo-random sequence generator, and wherein the decoding the at least one first monitoring information further comprises causing the apparatus to: carry out a dithered dequantization process using the at least one sec- ond initial value in generating at least one second dithering value by the second pseudo-random sequence generator.

18. The method of claim 16 or 17, further comprising:averaging the at least one metric estimated during one or more channelstatus indication cycles.

19. The method of any preceding claim 16 to 18, wherein the resem-blance evaluation is carried out by using squared generalized cosine similarity.

20. The method of any preceding claim 16 to 19, wherein the firstpseudo-random sequence generator and the second pseudo-random sequence generator operate in a fashion that produces at least substantially same sequence when the at least one first initial value and the at least one second initial value are at least substantially same.

21. The method of any preceding claim 16 to 20, wherein the at leastone downlink channel transmission parameters comprise a modulation scheme, a code rate, a number of transmission layers, and / or multiple-input and multiple- output precoding.

22. The method of any preceding claim 16 to 21, wherein the at leastone channel state information reporting parameter comprises a compression level, a quantization level, a limit on the number of compressed bits per a signalling mes- sage, and / or an indication of a compression method.

23. The method of any preceding claim 16 to 22, further comprising:indicating the at least one first initial value to at least one user device asa part of channel state information reporting configuration message, or using a dedicated signalling message.

24. The method of claim 23, wherein the message further comprises in-formation on quantization for the at least one first monitoring information, number of bits in the at least one first monitoring information, information on the duration of the one channel status indication cycle, and / or periodicity of the one or more channel status indication cycle.

25. A method, comprising:obtaining, by a user device, at least one first initial value for monitoring performance of data compression, wherein the at least one first initial value is ob- tained in a manner providing at least substantial similarity with at least one second initial value used by a network node in the monitoring performance of the data compression, determining at least one monitoring information, wherein the at leastone monitoring information is determined by generating at least one vector repre- sentative of a plurality of channel state information estimates, by obtaining at leastone pseudo-random index value for selecting at least one element of the at least one vector, or obtaining at least one pseudo-random weight vector for determining at least one vector product with the at least one vector, wherein the at least one pseudo-random index value or the at least one pseudo-random weight vector are obtained by using the at least one first initial value as an input to a pseudo-random sequence generator, and transmitting the at least one monitoring information to the networknode for the monitoring performance of the data compression.

26. The method of claim 25, further comprising: carrying out, for the transmission, a dithered dequantization process forthe at least one monitoring information using the at least one first initial value in generating at least one dithering value by the pseudo-random sequence generator.

27. The method of claim 25 or 26, wherein the pseudo-random se-quence generator operates in a fashion that produces at least substantially same sequence with the at least one first initial value than a second pseudo-random gen- erator in the network node with the at least one second initial value.

28. The method of any preceding claim 25 to 26, further comprising:receiving, from the network node, a control message in association with the monitoring the performance of data compression, and transmitting an acknowledgement to the control message.

29. The method of claim 28, further comprising: receiving an indication on the at least one first initial value in the controlmessage.

30. The method of claim 28 or 29, wherein the control message is a con-figuration message or a reporting request message.

31. An apparatus, comprising means for performing the method accord- ing to any of claims 16 to 24 or 25 to 30.

Images