206 results about "Mixed mode" patented technology

The invention provides a training strategy determination method for a computing power cluster using chips of various models in a mixed mode and a related device, and relates to the technical field of graphic processing chips, computing power clusters, large models, model training and the like. The method comprises the following steps: determining each target model chip forming a computing power cluster and a corresponding quantity proportion; for each target model chip, determining performance evaluation of the target model chip in a preset performance type and weight distribution of each performance type under each alternative training strategy; for each alternative training strategy, performing weighting processing on the performance evaluation under each corresponding preset performance type in combination with each corresponding number proportion by using the weight distribution, matched with each target model chip, of the alternative training strategy, so as to obtain corresponding comprehensive training evaluation; and determining a target training strategy based on the evaluation parameter of each comprehensive training evaluation. By applying the scheme, the training strategy which is suitable for being used by the computing power cluster which mixedly uses multiple types of chips during training of a large model can be quickly determined at low cost.

The invention discloses a power coupling-considered capacity ratio estimation method for a network-constructing / network-following multi-machine hybrid system, and the method comprises the steps: building a GFL-GFM hybrid system model, combining VSG power coupling small signal modeling and a dynamic relative gain matrix (DRGA), and quantifying the coupling strength of active / reactive power to output voltage; the frequency domain impedance characteristic of the GFM converter is simplified, and an ideal voltage source series equivalent reactance equivalent model is constructed; deriving an analytical relationship between a generalized short-circuit ratio (gSCR) and a capacity ratio based on the model, and generating a ratio formula in combination with engineering boundary sub-scenes to form a ratio range suitable for multiple scenes; for coupling time-varying characteristics, a closed-loop mechanism of'mixed mode-coupling sensing-correction calculation-execution feedback 'is constructed, and the ratio is dynamically adjusted. According to the method, coupling quantization and multi-machine coordination control are deeply fused, the problem that the external characteristic model precision and the multi-working-condition adaptability are difficult to consider under complex environment coupling is effectively solved, and the engineering applicability of the hybrid fan system is enhanced.

The invention relates to the technical field of automatic driving and robots, and discloses a cross-modal feature fusion method and system. The method comprises the following steps: acquiring original images and original point cloud data respectively acquired by a vehicle-mounted camera and a laser radar; performing dimension pull-up processing based on the original image to obtain a first feature point cloud, and performing compression mapping processing based on the first feature point cloud to obtain a camera modal two-dimensional feature; obtaining a second feature point cloud corresponding to the original point cloud data, and obtaining a laser radar modal two-dimensional feature based on the second feature point cloud; generating a third feature point cloud according to the first feature point cloud and the second feature point cloud, and performing compression mapping processing based on the third feature point cloud to obtain a mixed modal two-dimensional feature; and generating a cross-modal fusion feature based on the camera modal two-dimensional feature, the laser radar modal two-dimensional feature and the mixed modal two-dimensional feature. According to the scheme of the invention, more accurate and reliable environment perception capability can be provided for intelligent driving.

The invention discloses a media file query and retrieval method and system based on an AI dialogue scene, the method combines a mixed mode of vector retrieval and keyword retrieval, realizes accurate retrieval of cross-modal data, and can process various media files including texts, pictures, videos, PPT and the like. According to the method, a double-round scoring mechanism of a large language model (LLM) is combined, firstly, the relevancy of file content is scored, then secondary scoring is carried out according to the matching degree of a file title and user query, and it is ensured that the file most relevant to query is called. In addition, the system can further utilize a user feedback mechanism to continuously optimize the retrieval precision, and the user experience is improved. Compared with the prior art, the method has the advantages that the accuracy and the efficiency of multi-modal data retrieval are remarkably improved, the limitation of a traditional AI dialogue system during multimedia file processing is solved, and the method is suitable for intelligent dialogue application scenes needing to quickly and accurately obtain various types of media files.

A computing system receives a video bitstream that comprises a current block within a current picture and one or more syntax elements. The current block is encoded in a subblock-based motion vector (MV) refinement mode using information from a first reference block and a second reference block. The system applies a MV refinement to generate a refined first MV indicating a first location of the first reference block and a refined second MV indicating a second location of the second reference block. Subsequent to applying the MV refinement, in accordance with a determination that (i) the first location is outside of a first set of reference boundaries and / or (ii) the second location is outside of a second set of reference boundaries, the system decodes the current block using a weighted average of respective values from the first and second locations.

The invention relates to the field of front-end technology, can be applied to the field of digital medical / financial science and technology, and discloses a front-end multi-tenant management method and device, electronic equipment and a storage medium. The method comprises the following steps: returning a security access token corresponding to a target tenant by using an authentication mechanism based on a token or third-party authorization according to a tenant registration request; when registration succeeds, role configuration data and permission configuration data of the target tenant are generated; when it is monitored that login is successful by using the security access token, creating a micro-front-end application and different micro-front-end modules of the target tenant according to the role configuration data and the authority configuration data; starting a micro-front-end application, and dynamically loading a micro-front-end module according to the role configuration data and the authority configuration data; and executing mixed mode rendering on the front-end module at the server side and the tenant side through the integrated front-end framework to obtain a front-end rendering page of the target tenant. According to the method, multi-tenant security isolation is realized, resources are loaded as required, and system performance and expansibility are guaranteed.

The invention relates to an equipment collaborative scheduling method in a sea-railway combined transport port mixed mode and electronic equipment, and belongs to the technical field of port engineering. The method comprises the following steps: by taking preset operation sequence connection, safe operation of a rail-mounted gantry crane and waiting time of each device as constraints, unloading a container onto a container truck by the rail-mounted gantry crane, and then transporting the container to a wharf by the container truck to wait for quay crane operation; a scheduling model is established according to the working condition of a transfer mode that the rail-mounted gantry crane stacks the containers needing to be stacked to the designated container positions of the railway storage yard; the scheduling model is solved based on the grey wolf algorithm introducing the chromosome uniform crossover strategy, a final equipment cooperative scheduling scheme is obtained, and the problem that in the prior art, research on the scheduling mode is single in actual work is effectively solved.

The invention discloses a direct current charging and discharging control system and method, and relates to the technical field of battery charging and discharging, and the system comprises the steps: monitoring the input and output voltage and current, the battery state parameters and the environment temperature in real time; judging whether the charging and discharging control mode is a charging mode, a discharging mode or a mixed mode according to the analysis result; if the charging mode is determined, the charging current is dynamically adjusted based on a self-adaptive closed-loop control unit; if a discharging mode is determined, dynamically distributing a discharging current based on a load grading strategy and the residual electric quantity of the battery; and if it is judged that the mode is the mixed mode, charging and discharging parallel control is carried out through the dynamic power distribution unit. The technical problems that in the prior art, a charging and discharging control strategy is single, complex and changeable working conditions cannot be adapted, and key parameters are difficult to accurately monitor in real time to achieve efficient battery management are solved, and the technical effects of improving the energy utilization efficiency and enhancing the equipment operation stability and reliability are achieved.

The invention discloses a large language model output quality guarantee method, device and system based on chain type deterministic reasoning verification, and belongs to the field of artificial intelligence safety, deterministic calculation and reasoning verification. The method comprises the following steps: classifying each reasoning step generated by LLM into a deterministic step (bT: deductive reasoning, verified facts and mathematical proof) or a non-deterministic step (bF: inductive reasoning, unverified references and creative guess), and accumulatively constructing a reasoning chain; chain certainty verification chain (chain) = foldr (AND, true, chain) and time complexity O (n) are carried out in each k steps; if the bF is returned, a chain pollution elimination theorem is applied to prove that once the chain contains non-deterministic steps, the whole chain loses deterministic guarantee; and a deterministic boundary report is generated, and a non-deterministic initial position is marked. The device comprises an inference classifier, a reference verifier, a mathematical checker and a chain verification engine, and all the modules can be independently realized (formalized verification / database query / rule engine / mixed mode).

The invention discloses a mixed mode anti-attack method and system based on a laser radar point cloud. The method comprises the following steps: acquiring an original point cloud; sequentially performing frequency domain transformation, disturbance and inverse transformation processing on the original point cloud to generate a disturbance point cloud; generating an initial antagonistic point cloud based on the disturbance point cloud; optimizing the initial antagonistic point cloud to obtain first antagonistic disturbance, wherein the first antagonistic disturbance comprises the following steps: constructing a target function based on a Hausdorff distance and a chamfering distance between the antagonistic point cloud and the original point cloud; obtaining a first antagonistic disturbance by minimizing the objective function, including: calculating gradient information in the current iteration; the step length is adjusted according to gradient information in the current iteration, the antagonistic disturbance in the next round of iteration is determined, and the first antagonistic disturbance is output after iteration is conducted for the preset number of times; and processing the original point cloud and the first adversarial disturbance based on shape prior guidance to generate an optimal adversarial point cloud so as to carry out adversarial attack. According to the method, the attack resisting effect and robustness are improved.

The invention provides an all-optical switching system and method based on AWGR and WSS, and relates to the technical field of optical communication, the system comprises a three-stage Clos network structure, the input stage is composed of N N * N WSS modules and used for receiving external optical signals, and the WSS represents a wavelength selection switch; the middle stage is composed of AWGR modules and WSS modules which are deployed in a mixed mode, the number proportion of the AWGR modules is adjustable so as to achieve balance of cost and performance, and AWGR represents an array waveguide grating router; the output stage is composed of N N * N WSS modules; the output port of the input stage WSS module is connected with the input port of the intermediate stage AWGR / WSS module, and the output port of the intermediate stage AWGR / WSS module is connected with the input port of the output stage WSS module. While the optical path blocking performance is ensured, the system cost and insertion loss are reduced, wavelength conflicts are reduced, and the balance between the cost and the performance is realized.

The embodiment of the invention relates to the field of storage device application, and discloses a flash memory management method, a storage control chip and flash memory device.The flash memory management method comprises the steps that based on the storage mode of each physical block in a super block, data operation is executed on the physical blocks in the super block; and when an error occurs in the operation process, converting the storage mode of the physical block so as to operate the physical block according to the converted storage mode. The storage modes of the physical blocks are converted when errors occur in the operation process, the physical blocks can continue to execute the data operation, the probability that the physical blocks are directly marked as bad blocks is reduced, the use cycle of the physical blocks is prolonged, the data operation is executed on the physical blocks in the super block based on the storage mode of each physical block in the super block, and the storage efficiency of the physical blocks in the super block is improved. According to the method and the device, fine management can be carried out on the super block in the mixed mode (namely, physical blocks in different storage modes exist in the super block), and the stability of data storage of the flash memory device and the durability of the flash memory device are improved.

A physical vapor deposition system may include an RF generator configured to transmit an AC process signal to a physical vapor deposition chamber via an RF matching network. A controller of the RF matching network receives the DC magnitude and phase error signals and varies an impedance of the RF matching network in response to the DC magnitude and phase error signals. The matching network operates in a first mode until a tuning dead-zone is determined. Once a tuning dead-zone is determined, the matching network operates in additional modes until the network is tuned. The controller uses a composite value of magnitude and phase error to drive the variable tuning and load capacitors. In some cases, a blended mode (representing multiple tuning algorithms concurrently) may be implemented as a single mode that weights across what would have been multiple modes and thereby tunes the network using a weighted blended mode.

The invention provides a self-adaptive noise source impedance adjustable EMI filtering device and method, and belongs to the field of electromagnetic interference processing. The device comprises an EMI filter, a coupling inductor, a balance winding, a balance impedance, a balance impedance adjusting module and a self-adaptive control module; the coupling inductor, the EMI filter and the balance winding form a six-port network, and the self-adaptive control module is used for converting the six-port network into a four-port network according to scattering parameters of the six-port network, constructing a mixed-mode common-mode parameter matrix, obtaining an initial common-mode reflection coefficient according to the mixed-mode common-mode parameter matrix, and outputting the initial common-mode reflection coefficient. And based on the initial common-mode reflection coefficient and a preset target reflection coefficient, obtaining a balance impedance adjustment direction and a balance impedance adjustment amplitude which meet preset requirements so as to control an impedance adjustment module to dynamically adjust the balance impedance. According to the invention, the balance impedance is optimized and adjusted in real time to cope with the dynamic change working condition of the circuit, and the EMI filtering performance reliability is improved.

The invention relates to an analog-digital hybrid SRAM (Static Random Access Memory) memory audio noise reduction method and system, and the method comprises the following steps: carrying out the differential sampling quantization processing of an input audio signal, and obtaining a multi-bit-width digital audio code stream; inputting the multi-bit-width digital audio code stream into a preset SRAM (Static Random Access Memory) memory chip to carry out mixed mode write-in operation to obtain an analog-digital mixed memory matrix; performing frequency domain feature extraction on the analog-digital hybrid storage matrix to obtain frequency domain extraction features; performing noise identification and suppression on the audio signal based on the frequency domain extraction feature to obtain a time-varying noise suppression coefficient; and performing audio signal recovery processing on the time-varying noise suppression coefficient to obtain a denoised output audio signal, thereby solving the technical problem that power consumption and hardware cost are greatly increased due to the fact that a traditional digital signal processing method generally depends on a high-precision analog-to-digital converter and a high-performance processor.

The invention is suitable for the technical field of mixed modal retrieval, provides a financial risk identification method based on multi-granularity mixed modal retrieval, and is suitable for the field of mixed modal retrieval. The method comprises the following steps: constructing a multi-modal retrieval library, respectively encoding image and text modal information by using CLIP and RoBERTa models, and aligning features through a learnable projection matrix to form a searchable financial knowledge base; a multi-granularity retrieval scheme combining Dual-Encoder and Cross-Encoder is adopted, a Top10 candidate answer is rapidly positioned through coarse granularity, and a high-relevance candidate set is accurately screened through fine granularity; and finally, through a multi-input weighted integration strategy, integrating multiple rounds of prediction results of the large model and carrying out weighted summation based on association degrees to generate a final risk prediction result. According to the method, limitation of a single data source is broken through, comprehensive utilization of cross-modal information is realized, information quality and retrieval efficiency are ensured, model uncertainty is remarkably reduced, a risk identification system with robustness and accuracy is formed, and scientific decision support is provided for financial institutions.

The invention discloses an APP interface security verification method and system in an iOS mixed mode, and the method comprises the steps: enabling an APP to initiate a request to a server side, obtaining a current security verification code and validity period information, and storing the security verification code and the validity period information in a data storage region of the APP; intercepting a native network request and a WebView network request of the APP to obtain a first network request; obtaining a security check code in a data storage area of the APP, and checking the rule to generate a security check parameter; modifying an interface domain name in the first network request, and respectively adding the security verification parameter to the corresponding first network request to obtain a second network request containing an encryption parameter; performing security verification on the second network request at the server side according to the verification rule, and returning a verification result to the APP; and executing corresponding operation according to the verification result. The invention relates to the technical field of safety verification, and solves the problem of potential safety hazards of an interface under a mixed mode architecture.

The invention belongs to the technical field of composite material structural mechanics and computational mechanics, and particularly relates to a composite material fan blade composite crack propagation simulation method, program and equipment and a storage medium. According to the method, key rotation and a failure judgment criterion based on a key deflection angle are introduced into the conventional state near-field dynamic model, so that the crack propagation behavior of the composite material under the shear or mixed mode load can be more accurately captured and simulated. According to the flat plate mapping method, the improved PD model suitable for the flat plate is successfully expanded and applied to the fan blade airfoil section, the problem of complex geometric modeling is solved, and the flat plate mapping method is suitable for complex geometric structures. According to the method, a near-field dynamics method is adopted, crack paths do not need to be preset or grid reconstruction does not need to be carried out, and complex expansion behaviors such as spontaneous bifurcation and merging of cracks can be naturally simulated; the PD model can be used for meticulously describing the mechanical behaviors and crack evolution of each layer and interlayer in the laminated plate, including intra-layer cracking and interlayer layering.

Methods, apparatus, systems, and articles of manufacture are disclosed for performing artificial intelligence-based sparse calculations based on mixed mode and dynamic coding. An example apparatus includes a memory, computer readable instructions, and processor circuitry to execute the computer readable instructions to: determine a hybrid sparse pattern of a selected layer of an artificial intelligence (AI)-based model, the hybrid sparse pattern having a sparseness ratio and a block pattern for the selected layer; in response to the sparseness ratio being above a threshold, reducing the sparseness ratio of the selected layer; and in response to the sparseness ratio being below the threshold, adjusting a block mode of the selected layer, the block mode of the selected layer corresponding to the accuracy ratio.

The invention provides a video editing method and electronic equipment. The method can be applied to the electronic equipment with image processing capability, such as a smart phone, a tablet computer and the like. When the video editor is used for editing the HDR video and the SDR video in a mixed mode, the electronic device can conduct format conversion on the HDR video and the SDR video based on the video brightness range corresponding to the video format needing to be output, the video format is unified so as to be compatible with the screen specification of the electronic device, and therefore the quality of the video output by the electronic device can be improved.

The embodiment of the invention relates to the technical field of converters, and discloses an active turn-off method, system and device for a hybrid commutation converter, and the method comprises the steps: obtaining the change rate of the current of a current bridge arm with time and commutation angle information in real time when commutation begins; when the change rate of the current of the current bridge arm along with time is abnormal, actively turning off a converter valve of the current bridge arm, and enabling the current to be subjected to phase commutation to the next bridge arm; and when the commutation angle reaches a set value, actively turning off the converter valve of the current bridge arm, so that the current is subjected to commutation to the next bridge arm. According to the embodiment of the invention, the commutation failure can be effectively resisted by detecting the current change rate; the natural commutation point effectively resists commutation failures under all working conditions and is a supplement of a current change rate criterion, the natural commutation point is turned off, so that the voltage and the current basically cross zero at the same time, and the reactive power demand is relatively low; the two turn-off methods are used in a mixed mode, and full-range commutation failure resistance can be achieved under small stress.

A mixed-mode operation method is applied to a DC power converter. The DC power converter receives an input voltage, and the input voltage is between a minimum voltage and a maximum voltage. The method includes steps of: operating the DC power converter in a SRC mode when the input voltage is greater than a lower threshold voltage and less than an upper threshold voltage; operating the DC power converter in an LLC mode when the input voltage is less than the lower threshold voltage; operating the DC power converter in the LLC mode when the input voltage is greater than the upper threshold voltage.

The invention provides a mixed-mode beat control method and system for a press in a continuous stamping production line, and relates to the technical field of industrial automation control, and the method comprises the steps: determining a cooperative control standard, an operation safety boundary and a control technical index through obtaining related data of the continuous stamping production line; on the basis, the initial phase interval of a sliding block of the target press switched to the discontinuous mode is adjusted, a target phase interval is obtained, and a data transmission channel among all the presses, the mechanical arm and the robot in the continuous stamping production line is established; first operation state data of all the presses, the mechanical arm and the robot are collected through the channel, so that the actual operation rhythm of the mixed mode of the presses in the continuous stamping production line is adjusted, the adjusted rhythm is obtained, the continuous stamping production line can operate according to the continuous mode after any press is switched to the discontinuous mode, and the continuous stamping production line can operate according to the continuous mode. And the self-adaptive coordination of the whole-line rhythm in the mixed mode of the press is realized.

A method and an apparatus for audio processing in a multi-view mode are provided. triggering, when a preset audio output adjusting event occurs, corresponding views to perform audio output by using matching sound output devices based on a policy that a first application uses a real sound output device and a non-first application uses a virtual sound output device, wherein the audio output adjusting event is an event causing the sound output devices used by the views to not match the policy, when a mixing mode is off, the first application is an application currently in an audio focus, and when the mixing mode is on, the first application is an application currently participating in mixing.

The invention relates to a fly-by-wire flight control system comprising a dissimilar design ACE and an ACE reconstruction method. The fly-by-wire flight control system mainly comprises a plurality of ACEs with dissimilar designs, a plurality of FCMs, a plurality of remote control electronics and a plurality of signal processing units. The non-similar ACE design can alleviate the risk of ACE common mode. When one or some ACEs are invalid or enter a direct mode, the fly-by-wire flight control system can reconstruct the ACEs of non-similar design, the reconfiguration capability of the system under the ACE failure is enhanced, the control surface is controlled as much as possible, and the influence of a mixed mode is reduced.

The invention relates to the technical field of unmanned site modeling, and particularly discloses an intelligent modeling method, device and system for an unmanned site digital twin model, and the method comprises the steps: carrying out the construction of the digital twin model of an unmanned site through the data of an unmanned site and the basic environment around the unmanned site; and then optimizing the unmanned station digital twin model based on deployment data of an actual scene corresponding to each grid in the established unmanned station digital twin model, namely, following a mixed mode of combining mechanism modeling and data-driven modeling, and taking a physical model as a framework. And the parameters of the digital twin model of the unmanned station are calibrated by using real-time data, and the precision and generalization ability are balanced, so that the construction accuracy and reliability of the digital twin model of the station are improved, and the real-time monitoring accuracy and reliability of the coverage area of the station in a complex environment are improved. Therefore, the accuracy and timeliness of task dynamic planning, reconstruction and bookbinding are higher, so that when tasks need to be executed, the control accuracy and efficiency of single task and multi-task cooperation in the site can be improved, the possibility of task planning again is reduced, and the task scheduling efficiency and accuracy are improved.

The invention provides an image rendering method and device, electronic equipment and a storage medium, and relates to the technical field of computers. The method comprises the following steps: for a target parent display object corresponding to a target frame image, in response to a starting operation of a preset batch hybrid switch for determining the target parent display object corresponding to a plurality of target display objects, obtaining a first hybrid mode parameter of the target parent display object; determining a first mixed mode parameter of the target parent display object as a target mixed mode parameter of each target display object, and setting the first mixed mode parameter to be invalid for the target parent display object; and performing rendering display on the target frame image according to the target mixed mode parameter of each target display object, thereby realizing rendering on the main screen according to the target mixed mode parameter of the target display object without creating a rendering target for the target parent display object, realizing batch rendering, reducing the number of times of state switching, and improving the display efficiency. And rendering performance is improved.

The disclosure describes systems and methods for mixed‑mode, mixed‑precision computation using a configurable compute engine for vector–matrix operations. A processing system selects an operating mode including at least a floating‑point mode or an integer mode for the compute engine. Operand storage and format preprocessing route differently sized element representations through shared lanes. In floating‑point mode, compute circuitry generates a mantissa product and a product exponent, and alignment applies power‑of‑two control through a selection network, including as part of a fused multiply‑accumulate path. Parallel adder trees feed a hierarchical output accumulator sized to the selected precision. Normalization operates at a wider intermediate precision, applies block‑level scaling with dequantization, and, in lower‑precision operation, performs groupwise normalization with an intermediate‑precision combine to preserve a constant compute‑block size. Implementations integrate the compute engine within a memory macro with double buffering or deploy the compute engine as a standalone accelerator.

This invention provides a remote key injection method, system, and POS machine. The method includes: sending a connection request to a key distribution host, the connection request including device capability information and / or key requirement information; receiving response information sent by the key distribution host, the response information including a first ciphertext obtained by encrypting a working key based on a first transmission key, or a second ciphertext obtained by encrypting the working key based on a second transmission key, or a third ciphertext obtained by encrypting the working key based on a first transmission key and a second transmission key; and decrypting the ciphertext to obtain the working key. This scheme is compatible with devices supporting TR34, devices supporting ECKA, and devices supporting mixed mode. The server can automatically select the most efficient or most secure protocol for interaction based on the algorithm capabilities reported by the device, greatly improving the efficiency, security, and flexibility of key distribution.

The present application discloses a communication data transmission method, device, equipment and storage medium, which relates to the field of information security technology, including: sending the first security alliance load determined based on the pre-quantum cryptographic algorithm, the post-quantum cryptographic algorithm and the hybrid mode recommended load to the responder through the initiator, so that the responder can send the second security alliance load; if it is a spliced ​​hybrid mode, the pre-quantum double certificate and the post-quantum double certificate are sent to the initiator through the responder, and if it is a nested hybrid mode, the pre-quantum and post-quantum hybrid double certificates are sent to the initiator through the responder; the identity authentication of the two communicating parties and the generation of the target key parameters are performed, and after the authentication is successful, the target key parameters are verified; if the verification is successful, the key exchange is performed based on the target key parameters and the post-quantum cryptographic algorithm, the session key is determined, and the communication data transmission is performed according to the security message protocol and the session key. The quantum security enhancement of the national secret Internet security protocol is realized.