276 results about "Information compression" patented technology

The invention discloses a multi-modal multi-scale retrieval enhancement generation method, system and device applied to external knowledge questions and answers and a medium. The method comprises question perception, multi-modal multi-scale query fusion coding, dense recall and answer generation. Analyzing a key query phrase from the question through a fine-tuned instruction language model, and accurately positioning a region of interest corresponding to the phrase in an image by using an open set visual positioning model; multi-source information is compressed and distilled into an optimal query vector through a deep fusion network integrating multi-head self-attention and an information bottleneck theory; executing a maximum inner product search to recall related knowledge; guiding the large language model to synthesize all information to generate a final answer; the system, the equipment and the medium directly perform feature fusion in the vector space based on the method, so that challenges such as information loss and cascading errors caused by a traditional normal form can be effectively dealt with, high correlation and high accuracy of retrieval knowledge are ensured, and accurate and reliable image-text questions and answers are realized.

The invention relates to the technical field of emergency communication, in particular to a Beidou short message-based multi-mode emergency communication terminal, which comprises a Beidou communication module, a multi-source positioning unit, an information coding processor and a main control unit, the Beidou communication module is used for realizing short message communication with a Beidou satellite and supporting a BDS-3 protocol; the physical SOS button is provided with a mistaken touch prevention structure and comprises a mechanical pressing assembly with a triggering pressure threshold value larger than 2N and continuous pressing time verification logic; the multi-source positioning unit is used for acquiring position information including longitude, latitude and height; the information coding processor is configured to compress the collected help-seeking information to be within the capacity limit of a Beidou short message and generate a standard message of an emergency communication protocol; a standby power supply switching circuit is arranged in the main control unit, the main control unit is used for coordinating the work of each module and managing a power supply system, and the power supply system comprises a replaceable lithium thionyl chloride battery and a solar charging panel; and the low-power-consumption main control unit controls the Beidou communication module to transmit information to a Beidou communication satellite.

The invention discloses a sparse dynamic 3D Gaussian splash method based on global-local feature extraction. According to the method, a global-local feature extraction module, an inter-frame feature flow aggregation network and depth prior are combined with a 4D pseudo pose supervision mechanism for cooperative work, and high-quality dynamic scene reconstruction and rendering are realized under the condition of a sparse camera visual angle. According to the method, the multi-view image information is compressed to the feature space for processing, so that the calculation complexity is reduced, and the geometric structure and the dynamic features of the scene are effectively reserved; the fused depth prior information and the 4D pseudo pose provide accurate geometric constraints, and the inter-frame feature flow aggregation network optimizes the time consistency of the dynamic scene. The method is particularly suitable for processing a complex dynamic scene under a sparse view angle, the sense of reality and the real-time performance of virtual view synthesis are remarkably improved, and a dynamic view with a more accurate geometric structure, higher time continuity and richer details can be generated.

The invention discloses a three-dimensional point cloud geometric information compression method based on implicit neural representation, and the method comprises the steps: constructing a trunk structure of an implicit neural network through a plurality of sine representation network layers which are connected in series, and introducing a variable-scale position coding mechanism on this basis, the method enables a network to obtain higher geometric reduction precision while keeping a compression ratio, and comprises the following steps: (1) inputting space coordinates of divided voxels into a position coding module with adjustable scale parameters; (2) feeding a coding result into an implicit neural network constructed by a network layer based on sine representation, and outputting the occupancy probability of the voxel through an activation function; (3) in a training stage, the model continuously optimizes parameters, so that the output probability distribution is highly consistent with a real occupied label; (4) after model training is completed, a method of combining an AdaRound second-order quantization optimization strategy and quantization perception training is introduced, network weight is finely adjusted, and quantization errors are reduced; (5) in a reasoning stage, judging whether the voxel is occupied or not according to a preset threshold value, and when the prediction probability exceeds the threshold value, regarding the voxel as occupied; and (6) all voxels judged to be occupied are aggregated, and reconstruction of the geometric structure of the point cloud is completed.

The invention discloses a federated learning image classification method based on time extension gradient cumulative average. Each client uses a local data set to train a local model to obtain gradient information of a current communication round, and uploads the gradient information to the central server after compression; the client continues to perform local training in a time period from the time when the client starts to upload the compression gradient information of the current communication round to the time when the server completes the compression gradient information aggregation of the current communication round; the central server side aggregates the compression gradient information to obtain average compression gradient information of the current communication round, and issues the average compression gradient information to the client side to update a local model of the (t + beta) th communication round; and repeating the steps until the training is completed, and inputting the to-be-detected image into the local model to obtain an image classification result. The method shows excellent performance under various data sets and different compression levels, has higher convergence speed and model precision, and solves the problem of low communication efficiency caused by network delay and bandwidth limitation in image classification of federated learning.

The invention discloses an efficient image analysis and question answering system for intelligent customer service, and aims to solve the commercialized problems of slow response and high cost when an existing AI customer service processes user picture questions. In order to improve customer service interaction experience, the system greatly reduces the amount of visual data input into a core language agent through an innovative visual information compression engine, for example, the amount of the visual data is compressed to one fourth of the original amount. By means of the key optimization, the intelligent customer service staff can respond to picture inquiry of customers in real time, and customer loss caused by waiting is avoided. Meanwhile, the system ensures high-precision identification of commodity details through a unique training strategy, provides reliable and accurate answers for customers, and successfully solves the problem that speed and precision are difficult to consider in customer service application in the industry. Finally, a practical solution with instant response and high reliability is provided for the fields such as e-commerce intelligent customer service and the like needing real-time interaction, and commercialized popularization of the advanced AI technology is powerfully promoted.

The invention discloses a system security access control method based on deep learning and dynamic risk awareness, which comprises the following steps of: generating a behavior state vector by collecting user identity, operation, equipment state and environment information, constructing an access behavior map, and establishing an access behavior map through a Hebbian learning rule; the method comprises the following steps: dynamically updating connection strength between behaviors in combination with historical behavior data, after receiving an access request, selecting an association path from a map, calculating an activation potential value, selecting an optimal path as an access decision basis by applying a Winner-Take-All selection mechanism, executing an access control operation, generating a risk mark and a behavior evolution mark, and establishing an access control strategy; the method is used for adjusting a selection threshold and an atlas weight. And finally, compressing and coding the related information, and recording the information in an access log. According to the method, intelligent judgment on the access behavior is realized by using deep learning and a dynamic risk perception mechanism.

The invention provides a multi-modal large-model long-sequence information compression retrieval method, and relates to the technical field of data processing, and the method comprises the following steps: receiving multi-modal long-sequence original data of a target patient; all modal data in the multi-modal long sequence original data are uniformly mapped to the same hidden space through a pre-trained multi-modal large model, and a unified semantic representation sequence of semantic association between modals is obtained; performing time sequence feature analysis on the unified semantic representation sequence, regarding feature points in the semantic representation as a spatial point set, determining boundary distribution by constructing convex hulls of feature clusters, identifying key feature clusters in the semantic representation, and determining a key time sequence interval based on a convex hull distribution mode of the key feature clusters. According to the method, the problems of weak cross-modal semantic association, easy loss of key time sequence information and low compression and retrieval efficiency in the existing multi-modal long sequence data retrieval are solved.

For example, a processor may be configured to generate compressed radar information by compressing radar values in a plurality of data bins of at least one radar processing dimension, the at least one radar processing dimension including a range dimension. For example, the processor may be configured to generate the compressed radar information by quantizing a plurality of normalized values corresponding to the radar values in the plurality of data bins. For example, a normalized value corresponding to a radar value in a data bin may be based on a normalization of the radar value with respect to a plurality of radar values in the data bin. For example, the processor may be configured to store the compressed radar information in a memory.

A communication method and related devices, in which a first communication device may transmit the second channel information, enabling a second communication device to communicate based on the second channel information. Wherein the second channel information is determined according to the first channel information, the time information of the first channel information and the characteristic information of the historical channel information, namely, the second channel information can reflect the time domain correlation corresponding to the time information of the first channel information and can reflect the time domain correlation corresponding to the time information of the historical channel information. And one or more of time domain correlation, space domain correlation and frequency domain correlation corresponding to the characteristic information of the historical channel information can be reflected. Therefore, the accuracy of the second channel information can be improved by utilizing various correlations, so that the performance of data transmission based on the second channel information is improved. In addition, the scheme can be applied to a channel information compression feedback scene, and the overhead of channel information feedback can be reduced.

The invention provides a sparse adaptation circuit and a network-on-chip sparse data processing method, the sparse adaptation circuit is embedded between a network interface unit (NIU) and a router of an NoC system, and the sparse adaptation circuit comprises an input processing module, a compression engine and an output processing module; the sparse detection unit is used for performing real-time zero value detection on the data stream received from the network interface unit NIU to generate sparse identification information; the compression engine is coupled with the sparse detection unit and is used for carrying out dynamic compression coding on the data stream based on the sparse identification information; and the output interface is used for sending the processed data flow to a router in a format conforming to the network-on-chip protocol. In the network-on-chip, a hardware method is adopted to identify a data packet containing a large number of zero values in AI calculation and compress the data packet into a data packet only containing valid data, so that the problems of low transmission efficiency, waste of power consumption and the like caused by the fact that on-chip bandwidth is occupied by invalid data are avoided.

An information processing method and apparatus, device and medium, the method comprising: receiving first interaction guidance information for an artificial intelligence model, the first interaction guidance information being used for guiding the artificial intelligence model to respond; a compression rule is obtained, the first interaction guide information is compressed into second interaction guide information based on the compression rule, and the number of lexical elements included in the second interaction guide information is smaller than that of the lexical elements included in the first interaction guide information; and sending second interaction guide information and guide additional information related to the compression rule to the artificial intelligence model to guide the artificial intelligence model to generate first response content based on the compression rule and the second interaction guide information. According to the method, in the interaction process of the client and the artificial intelligence model, the model input length and lexical element consumption can be reduced, the context window limitation in the interaction process is relieved, and the transmission efficiency of the interaction process is improved.

The invention provides a binocular video compression method based on parallax compensation and deep learning, and the method comprises the steps: firstly obtaining training data, carrying out the preprocessing, and then constructing a binocular video compression network; the binocular video compression network comprises a motion estimation network, a motion information compression network, a motion compensation network, a binocular video frame information compression network and a frame reconstruction network, then carrying out supervised training on the binocular video compression network, and carrying out video compression processing by using the trained binocular video compression network; according to the method, a binocular feature interaction model and a binocular entropy model are arranged in a motion information compression network and a binocular frame information compression network, the binocular feature interaction model aligns binocular visual angle features by using a mixed parallax compensation mechanism in the encoding and decoding process, and the binocular entropy model performs accurate probability distribution modeling through autoregression probability estimation. According to the method, the compression performance can be remarkably improved while the reconstruction quality is guaranteed, and the method is suitable for scenes such as virtual reality and automatic driving needing efficient binocular video processing.

The various implementations described herein include methods and systems for encoding video. In one aspect, a method includes receiving a mesh with polygons representing a surface of an object; detecting a first symmetric region in the mesh that includes a first symmetry line to divide the first symmetric region into a first partition and a second partition. The method includes recursively determining whether one of the first partition or the second partition includes a second symmetric region until no symmetric region is detected in both the first partition and the second partition. The method includes in response to detecting the second symmetric region within one of the first partition or the second partition: determining a second symmetry line within the second symmetric region to divide the first or the second partition into a third sub-partition and a fourth sub-partition; and compressing information of the third sub-partition, the second symmetry line and the first symmetry line into a bitstream.

The invention provides a portrait animation generation method and system based on a reversible encoder and dynamic optimization, and relates to the technical field of video portrait reconstruction, high-precision portrait animation generation is realized through multi-module collaborative innovation, multi-scale appearance features are extracted by using a VAE encoder, identity vectors extracted by ArcFace and global appearance embedding of CLIP are combined, and the portrait animation generation method and system based on the reversible encoder and dynamic optimization are realized. Constructing a multi-modal characteristic system; a high-precision attitude estimation algorithm is adopted to extract fine joint points, and an error between a generated attitude and an original attitude is calibrated in real time through a dynamic optimization strategy, so that the attitude driving accuracy is remarkably improved. A reversible residual block is designed to realize lossless two-way transmission of features, and the problem of information compression of a traditional encoder is avoided; a high-frequency detail injection mechanism is introduced, high-frequency features such as facial textures are extracted through a pre-training model and fused with shallow features, an attention mechanism guided by identity features is introduced based on the generation process of a diffusion model, and the consistency of character identities in the generation process is ensured.

The invention belongs to the technical field of landslide evaluation, and particularly relates to a modeling method of a landslide susceptibility evaluation coupling model fusing an auto-encoder and a feedforward neural network, and the method comprises the steps: obtaining slope unit data and historical landslide data, carrying out the preprocessing, extracting landslide-related evaluation factors, carrying out the correlation test and correlation analysis, and obtaining a landslide susceptibility evaluation coupling model. Screening out evaluation factors for subsequent evaluation; constructing a landslide data set, balancing samples by adopting a GPU acceleration oversampling technology, and dividing a training set and a test set; constructing a landslide susceptibility evaluation coupling model based on the auto-encoder and the feedforward neural network; training the landslide susceptibility evaluation coupling model by using the training set and the test set to obtain optimal parameters; and outputting a landslide susceptibility distribution result by using the trained landslide susceptibility evaluation coupling model. According to the method, feature dimension reduction and redundant information compression can be effectively realized, and the evaluation precision and generalization ability of the model are improved.

The invention discloses a method and a system for protecting privacy of source codes in memory error detection, belongs to the technical field of computer software security, and particularly relates to a technology for protecting confidentiality of the source codes in a third-party memory error analysis process of software. According to the method, a debugging information minimization processing system DIREDUCER is constructed, a selective deletion and type minimization technology is utilized, on the premise that complete source codes are not exposed, necessary debugging information used for memory error detection is effectively reserved, and dual guarantee of source code privacy and memory error detection effects is achieved. According to the method, the debugging information in the non-stripping binary file is compressed to be within 10% of the original volume, and the recognition capability of an analysis tool on the problems such as memory leak, buffer overflow and overhanging pointers is not remarkably reduced. Experimental results show that the method has good adaptability and expandability in actual deployment, is suitable for various debugging tools and binary analysis frameworks, and has wide application prospects and practical values.

A user equipment (UE) includes a transceiver and a processor configured to transmit, via the transceiver to a network, UE capability information that corresponds with performance monitoring for artificial intelligence (AI) model-based channel state information (CSI) compression at the UE. In accordance with the UE capability information, the processor is configured to receive, via the transceiver and from the network, configuration and activation instructions for performing the AI model-based CSI compression and a network configuration on assisted information for the performance monitoring. The processor is configured to monitor performance of the AI model-based CSI compression in accordance with the received UAI configuration, and based on the monitored performance of the AI model-based CSI compression, transmit, via the transceiver to the network, a preference of the UE for the AI model-based CSI compression.

The invention relates to a method for realizing semantic chain retrieval enhanced generation processing for knowledge questions and answers of long documents. The method mainly comprises two parts of knowledge base construction and user knowledge questions and answers. User knowledge questions and answers are constructed based on a knowledge base, and knowledge base construction relates to long document preprocessing, semantic chain type storage, information compression and knowledge extraction and storage. When a user performs knowledge questions and answers, a retrieval vector is formed through fine-grained semantic coding, and then related document vectors and knowledge maps are matched in a knowledge base through vector similarity calculation. If the length of the document exceeds the context scale of the model, document information after static information compression is indexed through information compression, whether dynamic compression is carried out or not is evaluated, and document fragments related to tasks are returned through dynamic compression. Otherwise, directly returning the related original document as the background knowledge of the generator. And the model generator generates an answer based on the related document information / information after information compression, the knowledge graph information, the user question and a model prompt.

A hierarchical personalized federal learning method based on cross-client prototype clustering relates to the field of artificial intelligence and distributed machine learning, and comprises the following steps: a client maps a local sample to a semantic prototype space for information compression, and generates a category prototype matrix; the server receives the prototype matrix uploaded by each client, dynamically clusters and aggregates the prototype matrix, constructs a multi-level knowledge structure including a cluster-level expert model and a global semantic prototype, and maintains the global semantic prototype through index moving average; the client receives the cluster-level expert model, the global model parameters and the global semantic prototype from the server, and carries out a new round of local training on the basis; the client integrates local data and multi-level knowledge issued by the server to optimize three types of loss updating models; through loop iteration, the local model of the client gradually realizes personalized adaptation and global alignment. According to the method, the communication efficiency, the privacy protection safety, the model convergence speed and the generalization ability are improved.

The invention provides an interpretable safe longitudinal federal representation learning method and device for multi-modal big data. According to the method, for the problem of target domain modal data scarcity, in the target domain modal representation learning process, an attention mechanism is used for supplementing information of a source domain modal into a target domain modal, and an algorithm framework is established under a longitudinal federated learning framework, so that on one hand, it is guaranteed that data of a source domain and data of a target domain are not locally output, and the algorithm framework is established under a longitudinal federated learning framework; therefore, the data security is improved; on the other hand, the problem of insufficient modal data of the target domain is solved, and the performance of downstream tasks is improved; besides, in the process of constructing the loss function, an information bottleneck theory is introduced, redundant information between input and intermediate representation is minimized, and related information between representation and a target task is maximized, so that efficient information compression and feature extraction are realized, and the interpretability of extracted representation for downstream tasks is improved.

The invention discloses a GPU fault detection and processing method and device, equipment and a storage medium, and relates to the technical field of cloud computing, and the method comprises the steps: determining the position information of a video memory abnormal process based on the context switching frequency and verification error count of a GPU; compressing an association result obtained by dynamically associating each application performance index with the processor information of the GPU to obtain compressed data, and performing stream processing, time sequence prediction and twin mapping operation on the compressed data to obtain a detection result; processing the position information by using a target monitoring rule generated based on historical fault data and based on the target knowledge graph and the detection result to obtain a to-be-processed scheme; and generating a snapshot strategy by using an early warning signal generated based on the regional temperature distribution data of the GPU and the depth of the task queue, and processing the to-be-processed scheme based on the snapshot strategy by using the standby computing node to obtain a target scheme. Therefore, the GPU detection efficiency can be improved.

This invention discloses a Tai Chi movement recognition method based on multi-rate compressive convolution. First, electromyographic (EMG) data of Tai Chi movements is collected and preprocessed. Then, temporal features of the source signals are extracted to construct a feature matrix. A multi-rate compressive convolution algorithm is used to extract cross-channel compressed feature sequences, which are then fused to obtain a multi-dimensional fused feature matrix. Temporal relationship features are then obtained through a long short-term memory (LSTM) network. Finally, a fully connected layer is used to obtain the movement recognition result, thus completing the Tai Chi movement recognition. This invention achieves multi-scale cross-channel feature extraction of EMG signals through multi-rate dilated convolution, overcoming the limitation of ordinary convolution which can only extract spatial relationships between adjacent channels. It can compress the information contained in EMG signal sequences of arbitrary sizes into a one-dimensional compressed feature sequence to extract highly fused features, reducing data redundancy, the number of parameters, and model computation, thereby improving the accuracy of Tai Chi movement recognition.

The invention discloses an image subgradient constrained dual-camera snapshot spectral imaging system and an imaging method, and relates to the field of spectral imaging, in particular to an image subgradient constrained dual-camera snapshot spectral imaging method. The method aims at the technical bottleneck that a traditional spectral imaging system cannot be suitable for a dynamic scene and a low-illumination environment due to low time resolution and limited light utilization rate, and the problem that an existing SCSI system is poor in reconstruction accuracy of a decoding algorithm due to the fact that the compression ratio of space-spectral information is too high or information complementarity is insufficient. The method comprises the following steps: 1, obtaining a preprocessed broadband filter array RAW image and an SD-CASSI compressed encoding image Y; 2, obtaining a demosaiced multiband image; 3, obtaining a coarse estimation spectrum image; 4, calculating the subgradient of the three-dimensional total variation function f3DTV (.) at the roughly estimated spectral image, wherein the subgradient represents the spatial characteristics of the spectral image; and 5, iteratively solving a spectral image recovery problem by using an optimization algorithm to reconstruct a spectral image.

A method implemented in a wireless transmit and receive unit (WTRU) may include receiving one or more reference signals; a first codebook-based precoding matrix is selected based on the one or more reference signals, and a second precoding matrix is determined based on the one or more reference signals. The method may include determining a differential precoding matrix based on the first codebook-based precoding matrix and the second precoding matrix. The method may include determining a compressed differential precoding matrix based on the differential precoding matrix. The method may include transmitting feedback information indicating the compressed differential precoding matrix.

The application belongs to the field of biomedicine and particularly relates to a TB-level biomedical three-dimensional image data large data format archiving method, which comprises the following steps: calculating TB-level image sequence slice information, dividing the image sequence into a plurality of three-dimensional local blocks; quickly reading the three-dimensional local blocks in a serial reading and parallel decoding mode; cutting each three-dimensional local block to obtain a plurality of three-dimensional small blocks; compressing each three-dimensional small block by using the information of the z-axis continuous change of the three-dimensional small block; continuously writing the compressed data into a corresponding container to obtain 0-level large data format data; further downsampling, compressing and continuously writing the local blocks into a corresponding 1-level large data format container; and obtaining 2-level large data format data in the same way. The application directly and quickly reads the local blocks, compresses and makes TB-level biomedical three-dimensional image large data format by using the z-axis continuous information, has the advantages of low cost, strong practicability, simple operation and the like, and can be calculated in parallel to further accelerate the archiving.

The invention discloses a signal detection method and device, and relates to the field of communication. The method comprises the following steps: a centralized processing unit generates first compression auxiliary information corresponding to a first distributed processing unit according to global historical channel information, and sends the first compression auxiliary information to the first distributed processing unit; and the first distributed processing unit compresses a locally received first signal and a first demodulation reference signal according to the first compression auxiliary information to obtain a first compressed signal, and sends the first compressed signal to the centralized processing unit. And the centralized processing unit performs signal detection according to the first compressed signal. According to the invention, the distributed signal processing architecture can maintain low flow transmission requirements and processing time delay in a compression mode, and the signal detection performance is remarkably improved.

Various aspects of the present disclosure generally relate to wireless communication. In some aspects, a first wireless communication device may receive configuration information associated with uplink data compression, the uplink data compression being associated with at least one of multiple compression dictionaries or multiple compression buffers. The first wireless communication device may compress, in accordance with the configuration information, uplink data to generate compressed uplink data via at least one of: one or more compression dictionaries of the multiple compression dictionaries, or a compression buffer of the multiple compression buffers, and the one or more compression dictionaries or the compression buffer being selected in association with one or more traffic parameters of the uplink data. The first wireless communication device may transmit an uplink communication that includes the compressed uplink data. Numerous other aspects are described.

This invention discloses a translation method and system based on structured segmentation and contextual memory, comprising: performing multi-dimensional vectorization representation of the long text to be translated to establish a full-text semantic feature mapping space; adaptively decomposing the long text to be translated to obtain a tree-like hierarchical structure; compressing key information of semantic units, extracting core semantic feature vectors, and storing them in a globally dynamically updated contextual memory pool; generating an enhanced prompt information sequence based on the core semantic feature vectors and the hierarchical attributes of the current semantic unit; inputting the enhanced prompt information sequence into a large-scale language model to generate a candidate translation sequence; performing boundary verification and smoothing on adjacent candidate translation sequences to output a complete structured translation. This invention improves the automation level of long text translation by systematically optimizing the processes of text deconstruction, contextual memory, constraint generation, and global verification, ensuring the structural integrity, logical coherence, and semantic fidelity of the translation.

Disclosed in the present application are a data transmission method and apparatus. The method comprises: a transmitter device generating and transmitting first signaling, which comprises information of a base matrix, indication information of the base matrix, information of a compression coefficient matrix, and indication information of the compression coefficient matrix, wherein the base matrix comprises at least one base vector, the base vector corresponds to a first vector in an original data matrix, second vectors in the original data matrix other than the first vector form a non-base matrix, and the base matrix performs data compression on the non-base matrix to obtain the compression coefficient matrix; and a receiver device receiving the first signaling, and obtaining the original data matrix on the basis of the first signaling. The present application provides a method for reducing the overheads of dictionary data transmission during dictionary compression, and proposes how to perform standardized transmission on data and information generated during the process of the method, thereby ensuring the reliability of a data decompression process.