183 results about "Expert group" patented technology

A scalable platform for orchestrating networks of specialized AI multi-agent networks that enables secure collaboration through token-based protocols and real-time result streaming with advanced dynamic chain-of-thought pruning. The central orchestration engine manages domain-specific agents, implementing sophisticated multi-branch reasoning with contribution-estimation layers that evaluate each agent's utility using Shapley value-inspired metrics. The system employs information-theoretic and gradient-based surprise metric to guide memory updates and dynamic reasoning expansion, preventing local minima stagnation while preserving valuable insights through adaptive forgetting mechanisms. The platform unifies Monte Carlo tree search with contribution-aware estimation to detect high-synergy expert combinations while maintaining privacy through partial data approaches. It scales across distributed computing environments, enabling complex collaborative tasks like materials discovery, product engineering and manufacturing process design, biomedical research, and drug development. The system supports multi-party economic rewards through systematic contribution effort, cost and importance tracking, while standardized interfaces manage security, privacy, and policy constraints across heterogeneous agents.

The invention discloses a hybrid expert network training method, device and equipment and a computer program product, experts in a hybrid expert layer are equally divided into a plurality of groups and equally distributed to a plurality of pieces of equipment, and different pieces of equipment share the same routing module. In the training process, for an input hidden layer sequence of the mixed expert layer, the score of each path of experts is calculated through a routing module, for each expert group, activation experts in the group are selected according to the scores of the experts in the group, the number of the activation experts in different groups is the same, and each expert group processes the complete hidden layer sequence. And processing the input hidden layer sequence by the activation experts in each expert group to obtain the output of the activation experts, and obtaining the output of the hybrid expert layer based on the output of each activation expert. According to the method and the device, absolute balance of calculation among the devices is ensured, and the effect of the model is not influenced by drop-pad (namely data discarding and 0 filling).

The efficient generative task reasoning acceleration method based on the hybrid expert network comprises the steps that context information of related generative tasks is obtained, context data is preprocessed, and it is ensured that the data is suitable for model input; constructing an efficient reasoning model based on the hybrid expert network, and determining a reasoning process of the efficient reasoning model based on the hybrid expert network; training an efficient reasoning model of the hybrid expert network, optimizing model parameters, and storing an optimal model structure; and generating a reasoning result by using the optimal model structure. In a model improved by the method, an expert network is composed of a plurality of independent experts, and each expert is responsible for processing different input characteristics. The gating network dynamically selects which experts participate in the calculation according to input characteristics, and assigns an expert to each input by calculating a probability distribution. In this way, the gating network and the expert network are closely matched, it is ensured that only the most suitable expert is used, and therefore the calculation efficiency and the model performance are improved.

The invention provides a dynamic efficient routing method and device for a hybrid expert large model, and the method comprises the steps: obtaining input data and a pre-trained hybrid expert large model which comprises a distributor module and a router; dividing the input data into a plurality of lexical elements, and determining the optimal expert number of each lexical element by using a distributor module; and selecting an expert combination for each lexical element according to the optimal expert number by using the router, and routing each lexical element to the corresponding expert combination. According to the method, the number of activated experts can be adaptively adjusted according to the difficulty and importance of each lexical element, the mechanism not only improves the utilization efficiency of computing resources, but also further improves the performance of the model through refined resource allocation.

The invention provides a Zip-MoE model grouping hybrid expert layer-based Chinese and English speech recognition method and system, a Zip-MoE model comprises six encoder blocks, a Bypass module is included between every two encoder blocks, and the weights of the output of the previous encoder block and the output of the current encoder block are learned; the first three encoder blocks are of a standard Zipform structure; the last three encoder blocks adopt a Zipform-MoE structure containing a grouped hybrid expert layer, and the grouped hybrid expert layer is used for replacing the last feed-forward network of the Zipform structure; the grouping mixed expert layer comprises a Chinese expert group, an English expert group and a language router, and each expert group is composed of a plurality of expert networks and is provided with an unsupervised router. The problem of language confusion is relieved, different time delay streaming scenes can be adapted, the number of experts is flexibly expanded, pre-training is not needed, and the Chinese and English recognition efficiency is greatly improved.

The invention relates to the field of intelligent automatic driving, in particular to a knowledge-driven end-to-end automatic driving method based on a sparse expert mechanism and a diffusion model. Comprising the following steps: S1, sensing information processing and state coding; s2, sparse expert module construction and multi-task training; constructing a sparse expert module composed of a plurality of experts, and obtaining a reusable driving skill through multi-task behavior cloning training; s3, generating a diffusion strategy network and an action sequence; and on the basis of a diffusion model, a future multi-step control action sequence is generated from the current state condition, and a continuous and stable driving decision is formed. And S4, a continuous learning and task migration mechanism. According to the method, a combinable and explainable modular driving knowledge structure is constructed, so that the strategy modeling capability is remarkably improved; a diffusion generation mechanism effectively improves the smoothness and stability of the decision process; and a continuous learning and task migration mechanism of structural decoupling improves the long-term adaptability and deployment efficiency of the system.

The invention provides a DDD architecture service code review and evaluation method based on a multi-expert LLM, and belongs to the technical field of industry intelligence and the field of software process quality assurance. Multi-expert large model cooperation, token polling transmission and specific scene domain drive design (DDD) architecture analysis are adopted; in the same submitted code group, different dominant large models can be tried to serve as dominant large models, and final code review is performed to optimize an access condition of a suggested number. And through a multi-expert large model cooperation mechanism, the environment problem easily generated by a single large model is avoided, and the review stability is improved. The cooperation efficiency between large models is improved through a token polling mechanism of a large model expert group, fair speaking of each large model is ensured, an online state detection mechanism of a subordinate large model can ensure that tokens are transmitted to currently available large models, the situation that the overall review process is affected by individual large models is avoided, and the review efficiency is improved. And the business code review efficiency of the system is effectively improved.

The invention relates to an optimization method of a hybrid expert system, computer equipment, a readable storage medium and a program product. A plurality of experts contained in the MOE are deployed in a plurality of artificial intelligence chips in groups, and the method comprises the following steps: carrying out routing calculation on an original input tensor to obtain a routing calculation result; determining input element grouping information corresponding to each expert based on expert index information and expert weight information in the routing calculation result; taking the expert dimension as a parallel dimension, executing rearrangement operation for the original input tensor in parallel based on the input element grouping information, taking a rearrangement result as input information of an expert, executing matrix multiplication and accumulation operation, and taking the expert dimension as the parallel dimension, and based on the input element grouping information, executing anti-rearrangement operation of the matrix multiplication and accumulation operation result in parallel to obtain a final operation result. By adopting the method, the MOE reasoning performance can be improved.

The invention relates to a mine safety early warning emergency management decision and linkage response system based on'scene-quick response chain-response ', and belongs to the technical field of mine safety early warning emergency management decision and linkage response. The system comprises a multivariate information acquisition and input module, a multi-chain data fusion module, a disaster situation analysis and evaluation display module, an emergency knowledge platform based on a knowledge graph, a command scheduling and emergency capability evaluation display module, an emergency information intelligent receiving and issuing module, an emergency drilling and training module and the like. And an emergency management closed loop based on knowledge, expert group intelligence, an emergency process and an intelligent decision rule set is realized. According to the invention, the scientificity, compliance and timeliness of mine safety emergency management decision and linkage response can be effectively improved, and the mine safety service guarantee level and the working efficiency are improved.

The invention discloses a general steganalysis detection method for a spatial domain and a JPEG domain, which is characterized in that an SRM filtering kernel initialization branch and a randomly initialized convolution kernel branch are mutually fused through a well-designed bidirectional fusion module, the fused complementary information promotes the effective cooperation and learning of the two branches, the noise residual error extraction capability of the double branches is enhanced, and the noise residual error extraction efficiency is improved. Therefore, the signal-to-noise ratio of the preprocessing stage is improved. A multi-level feature fusion module is introduced in feature extraction, feature maps of different levels are weighted through a space attention mechanism, the feature maps of different levels are integrated by down-sampling to the same scale, and weak steganography features are reserved. In multi-scale attention pooling (MSAP) in a classifier, convolution kernels of different sizes are utilized to extract multi-scale features to enhance feature expression, then gated channel transformation (GCT) is utilized to control weight distribution among channels, and finally, style pooling of global standard deviation pooling and global average pooling is combined to perform global pooling on a feature map, so that the statistical magnitude is improved.

The invention provides an expert knowledge-driven large-model customized data processing method and related equipment. The method comprises the following steps: inputting acquired business data into a trained large model capable of dynamically customizing an expert sub-model; in the large model, based on the business data, determining a knowledge embedding vector; based on the knowledge embedding vector, determining the weight of each adapter expert, and customizing an expert sub-model; and processing the business data based on the expert sub-model. According to the embodiment of the invention, the business knowledge hierarchical structure is constructed by using the multi-dimensional business information and the expert knowledge based on the large-scale business data. In combination with a large-model efficient fine tuning technology and a mixed expert strategy, dynamic expert group adaptation is realized under the guidance of business data knowledge, so that different expert models are customized for different business data, automatic adaptation under multiple scenes is realized, and more professional data processing is realized.

The invention discloses an expert large model training method based on a mixed retention degree, and the method comprises the steps: inserting a mixed retention degree layer behind any layer of a selected basic large model, and obtaining an expert large model; setting a mixed retention degree layer; the mixed retention degree layer is composed of M experts, and each expert has a preset parameter; a routing module is introduced into the hybrid retention degree layer, experts are dynamically selected according to input features, and the randomness of routing is increased through Gaussian noise vectors; parameter values of the M experts are set through a setting strategy; and training the hybrid retention degree layer through a task-related loss function and a load balancing loss function, and in the training process, keeping the hybrid retention degree layer in a trainable state while keeping other parts of the expert large model in a frozen state. According to the invention, the designed load balances the loss, so that each expert is fully trained; and the difference of each expert is ensured from the source.

The invention provides a hydroelectric power supply insurance intelligent management method and system, and the method comprises the steps: generating a digital model based on the multi-mode process information of hydroelectric power supply insurance; based on the digital model, assisting the expert group to decide a hydroelectric power guarantee supply optimization strategy; and executing the hydroelectric power supply guarantee optimization strategy. According to the intelligent management method and system for hydroelectric power supply guarantee, various monitoring data in the hydroelectric power supply guarantee process are integrated into the digital model, the expert group is assisted to decide the hydroelectric power supply guarantee optimization strategy based on the digital model, and finally the hydroelectric power supply guarantee optimization strategy is executed. According to the method, the expert group can comprehensively master the real-time state of the power system in the optimization decision-making process, the response speed and the decision-making accuracy are improved, then, the expert personnel carry out optimization decision-making under the assistance of the system, and the optimization decision-making efficiency is improved.

The invention relates to the technical field of electric power task adjustment, and provides an electric power task model fine tuning method and system, terminal equipment and a storage medium, and the method comprises the steps: inserting a shared low-rank expert module in a pre-training model trunk, dynamically calculating the activation weight of each expert according to the input characteristics through a gating network, and carrying out the fine tuning of an electric power task model; and a sparse regularization constraint activation number is introduced. During training, expert output is subjected to weighted fusion to generate fine adjustment parameters, multi-task joint training is adopted, balance optimization is achieved by dynamically adjusting task weights, and during reasoning, only part of experts with the highest weights are activated, and sparse calculation is achieved. According to the method, a plurality of low-rank expert modules are inserted into a large model trunk, a trainable gating routing mechanism is introduced, the optimal expert combination weight is dynamically calculated according to semantic features of input task samples, and knowledge fusion, parameter multiplexing and reasoning sparse activation among multiple tasks are achieved.

The invention provides a multi-agent collaborative decision-making method, system and device, a medium and a product, and belongs to the field of business support, and the method comprises the steps: carrying out the registration and role distribution of multiple agents based on a member registration and election algorithm, and determining a round table member and a round table chairman; acquiring an external decision demand based on the round table chairman, selecting a round table member corresponding to the external decision demand as an expert group member, sending the external decision demand to the expert group member, and generating a proposal corresponding to the external decision demand by the expert group member; summarizing proposals fed back by the members of the expert group based on the round table chairman, distributing the summarized proposals to the members of the expert group, and voting the summarized proposals by the members of the expert group to obtain a voting result; voting results fed back by the members of each expert group are obtained based on the round table chairman, an optimal proposal is determined according to the voting results, and the optimal proposal is used as a decision scheme. The democratic votes are processed in parallel through multiple agents, the advantages of multi-agent decision making are exerted, and the decision making risk is reduced.

The embodiment of the invention provides a data processing method based on a grouped hybrid expert model, electronic equipment and a medium. According to the method, multiple pieces of input feature data are acquired, and for each piece of input feature data in the multiple pieces of input feature data, at least one target expert group matched with the input feature data is determined from multiple routing expert groups through a group routing unit; and determining at least one activation expert matched with the input feature data from each target expert group through an expert routing unit, and performing data processing on the input feature data through the sharing expert and the at least one activation expert to obtain a corresponding processing result. According to the method, the routing experts with the high adaptation degree with the input feature data are determined as the activated experts, distribution of the routing experts according to needs is achieved, the method only needs to carry out weight calculation on the routing experts in the target expert group, the calculation amount in the data processing process is reduced, the calculation cost is reduced, and the calculation efficiency is improved.

The invention discloses a hybrid expert model training optimization method based on expert calculation load balancing scheduling, and the method comprises the steps: firstly predicting current iteration global expert load data, and obtaining sampling data; and secondly, designing a multi-dimensional performance perception expert placement evaluation method, and realizing quantitative comparison and decision in a stage of searching an expert placement scheme by converting multi-dimensional communication and calculation into a unified time delay index. And then designing an expert placement strategy by adopting an expert grouping search algorithm of dynamic load feature perception and a matching algorithm of expert grouping and resource perception based on the sampling data and an expert placement evaluation method. Finally, on the basis of an expert placement strategy, an asynchronous communication strategy is adopted to implement expert and optimizer parameter scheduling, and training optimization is completed. According to the method, by improving the computing load balance and asynchronous communication concealment between the devices, synchronous waiting is reduced, and the training efficiency and the overall performance of the hybrid expert model are effectively enhanced.

The invention relates to a task execution method and device based on a hybrid expert model, equipment, a storage medium and a program product, and relates to the technical field of artificial intelligence. The calculation utilization rate and the calculation efficiency of the equipment can be improved. The method comprises the steps that the number of lexical units distributed to each expert in equipment is determined, the number of lexical units of each expert group in the equipment is obtained according to the number of lexical units of each expert, the expert group comprises one or more experts in the equipment, and an execution sequence corresponding to each expert group is determined according to the relative size of the number of lexical units of each expert group in the equipment, different devices are configured with the same number of expert groups, each expert group comprises the same number of experts, the expert groups in the same execution sequence in the different devices have the same relative size in the respective devices, and the communication task and the calculation task of each expert group are sequentially executed according to the execution sequence. The computing task of each expert group in the device is parallel to the communication task of another expert group in the device.

According to the time sequence risk control method and system based on the multi-level mixed experts, heterogeneous data in the medical field is fused and mapped to the unified representation space through multi-modal representation learning, and therefore data information is utilized more comprehensively. The system adopts a multi-level hybrid expert (MoE) architecture, tasks can be accurately routed to a more professional expert group according to patient characteristics through a hierarchical gating network, and the accuracy and resource utilization efficiency of the model are greatly improved. Meanwhile, through bidirectional interpretability and a causal inference mechanism, not only can the reason of a risk prediction result be explained, but also anti-fact analysis and specific intervention measure guidance can be provided, so that the model is converted from a pure prediction tool to an intelligent partner for auxiliary decision making.

The invention relates to the technical field of image data processing, in particular to an automatic driving model based on a mixed low-rank expert and a multi-domain adaptive fine tuning method, and the model comprises a general model base in a frozen state and a multi-domain adaptive component in a trainable state; the universal model base comprises an image encoder, a measurement information encoder and a trajectory planner; the multi-domain adaptation component comprises a mixed low-rank expert group and a domain sensing router; the mixed low-rank expert group is used for calculating a difference correction signal according to the input characteristics; and the domain sensing router is used for receiving the shared image features extracted by the image encoder, outputting probability vectors and dynamically activating corresponding experts in the mixed low-rank expert group according to feature distribution of input data. According to the method, plug-and-play type efficient adaptation of a single universal model to multiple downstream domains is achieved with extremely low calculation and storage cost, and the pain point of a traditional multi-domain adaptation method is perfectly solved.

The invention relates to the technical field of power grid image classification, in particular to an image classification method based on a semantic knowledge base in a smart power grid, and the method comprises the following operation steps: S1, data collection: 1.1, collecting image data from various devices (such as a monitoring camera and unmanned aerial vehicle inspection equipment) of the smart power grid; the images may include the appearance of the power equipment, the operation state (such as the oil temperature indication of a transformer and the on-off state of a switch), and the integrity of the line. 1.2, preliminarily preprocessing the acquired images, including unifying the format of the images (for example, converting the images into a JPEG (Joint Photographic Experts Group) format), adjusting the size (enabling the images to adapt to a subsequent processing algorithm, for example, uniformly adjusting the images into 512 * 512 pixels) and removing noise (removing salt and pepper noise in the images by using a filtering technology, for example, Gaussian filtering); according to the method, the feature extraction process can be guided through the semantic knowledge base, so that the model focuses on key features related to semantics, and the classification precision can be improved.

The invention relates to an intelligent bid evaluation expert extraction method and system in the technical field of bidding and tendering information management. The method comprises the following steps: S1, constructing an expert multi-dimensional portrait; s2, receiving review project information, analyzing and calculating relevancy between a project demand vector and each expert capability vector, and screening candidate experts; s3, performing intelligent scheduling and priority ranking according to the historical response rate, attendance rate and evaluation of the candidate experts; s4, carrying out credible evidence storage on key data of the extraction process by utilizing a block chain, and ensuring tampering prevention and traceability of the extraction process and expert information; s5, at the appointed time and place before bid opening, decrypting and reviewing expert sensitive information under supervision, and completing confirmation and notification of a final expert group; the system comprises an expert database management module, a semantic matching module, an intelligent scheduling module, a block chain evidence storage module, a security decryption module and a visual interface. According to the extraction method and the system thereof, the precision, the intelligence and the security of expert extraction can be realized.

The embodiment of the invention discloses a distributed video transmission architecture and method based on a swan gap system, a camera and equipment. In the video transmission framework, a video preview stream supporting acquisition of an MJPEG format is added to a camera service layer interface on the basis of an original camera driving framework of a swan gap system through an improved camera driving framework; processing the PipeLineCore layer and improving the PipeLineCore layer to output a video preview stream in an MJPEG (Multimedia Joint Photographic Experts Group) format; adding an MJPEG format definition in the improved distributed camera; during video transmission, a camera client of an improved distributed controlled end selects a video preview stream in an MJPEG format and sends the video preview stream to a camera server; the surface is directly called in the coding processing of the improved distributed controlled end; and calling an MJPEG decoder to decode the video stream in the decoding processing of the improved distributed master control end, and transmitting the decoded video stream to the virtual camera. The video transmission efficiency can be improved by improving the original video stream processing of the Nan Mongolian system.

The present invention discloses an intelligent rehabilitation assistance training system for spinal degenerative diseases based on deep learning. The system of the present invention includes a real-time classification module for traditional Chinese medicine guiding techniques videos based on deep learning and a video sequence division and evaluation module based on human skeleton representation; the former obtains two-dimensional human skeleton data as the training data of the learning model, conducts deep learning training to obtain a generalized deep learning model, and finally obtains the real-time frame classification result; the latter, according to the frame classification result, segments and corrects the skeleton sequences of the same category in real time, and compares and scores the segmented sequence segments with the skeleton sequence segments of the expert group videos of the corresponding category. The system of the present invention does not require the guidance and intervention of medical staff, enables patients to perform traditional Chinese medicine guiding techniques training by themselves at any time, is applicable to families and primary medical and health institutions, can relieve the pressure of medical staff, and improve the flexibility and accuracy of patients' rehabilitation training.

The invention discloses an MoE expert deployment system and method based on a wafer-level chip, and the system comprises a statistical module which is used for carrying out the statistics of expert co-activation probability distribution reflecting the cross-layer cooperative activation relation between experts in MoE and the communication demands of a calculation core on the wafer-level chip; and the clustering module is used for performing clustering based on the expert co-activation probability distribution and communication requirements by taking the minimization of communication traffic across physical regions and the minimization of intra-group communication load difference of expert groups borne by the physical regions as targets to obtain an expert grouping scheme of MoE and a physical region layout mapping scheme corresponding to wafer-level chips. According to the method, clustering can be carried out according to the cross-layer co-occurrence rule of the experts, and the experts subjected to high-frequency cooperative activation are constrained in the same physical region, so that a large amount of global communication is converted into local exchange, and the overhead of long-distance and high-delay cross-region communication is remarkably reduced. Meanwhile, load distribution is optimized in combination with communication requirements, and local hotspots are effectively avoided.

The invention discloses a joint optimization method for a joint photographic experts group (JPEG) quantization table and a truncation position based on rate distortion, relates to the technical field of JPEG coding optimization, and solves the technical problems that in a JPEG compression scheme, improvement of coding efficiency caused by joint optimization by using the quantization table and the truncation position is neglected, and the optimal quantization table and the optimal truncation position are difficult to find. According to the method, joint optimization is carried out on a quantization table and a truncation position of the JPEG by combining a JPEG compression method from the perspective of evolutionary computation, a strategy that the truncation position is fixed when the quantization table is optimized, a strategy that the quantization table is unchanged when the truncation position is optimized, an estimation method for designing fitness evaluation and a method based on rate-distortion optimization are respectively carried out, and the quantization table and the truncation position are optimized. The optimal quantization table and the truncation position can be effectively found, and the corresponding truncation position is optimized while multiple groups of quantization tables are generated in the optimization process; through joint optimization of the quantization table and the truncation position, image data can be compressed more effectively, redundant information is reduced, and therefore coding efficiency is improved.

The invention discloses an intelligent design system and method for a building structure. The method comprises the following steps: S1, constructing a building database and a rule database; s2, inputting a construction area and a construction target; s3, generating multiple groups of building designs; s4, generating a simulated building structure, and performing a simulated building structure test; s5, submitting an architectural design result passing the test to an expert group appraisal platform, and scoring by an expert group according to an assessment condition; and S6, outputting a final building design result. Model simulation and testing are carried out after the building scheme is generated according to the construction target, unqualified building design schemes can be filtered and processed through the testing step, the follow-up expert review data volume is reduced, scoring is carried out after expert review, and the efficiency is improved. And some basic qualified design schemes can be optimized according to suggestions of expert appraisal, so that the finally obtained design schemes can be enabled to be richer, and the efficiency of generating the construction design schemes of the scheme is further ensured.

The invention discloses an open vocabulary target detection method and system based on mixed expert fine tuning. According to the invention, a Grounding DINO model is used as a pre-training model, and a hybrid expert (MoE) module is respectively introduced into a feedforward neural network module in an original decoder and a cross attention module in a feature enhancement module, so that different expert combinations can process tokens of different semantic information, thereby enhancing the open vocabulary detection capability of the network. According to the method, the hybrid expert system is introduced into the Grouping DINO model for the first time, and experimental results on a data set show that the open vocabulary detection capability of the model obtained through fine tuning is improved. Meanwhile, the method disclosed by the invention can also be used as a plug-in to be added into other Transform structures, and excessive reasoning overhead cannot be generated.

The invention discloses a cross-task collaborative thinking chain distillation method, device and system, and a storage medium. The method comprises the following steps: S1, generating a thinking chain of a reasoning task data set; s2, performing difficulty sorting on the trained tasks according to the thinking chain; s3, performing orthogonal grouping training on the student model according to the thinking chain and the task difficulty; and S4, performing cross-task collaborative thinking chain distillation according to the LoRA expert combination. By adopting the technical scheme of the invention, negative migration is prevented by implicitly isolating parameters between unrelated tasks, and the reasoning ability of the small model is further improved.

The invention discloses a multi-task intensive prediction method based on mixed single-rank experts. The method comprises the following steps: collecting and arranging a multi-task intensive data set; dividing the data set into a training set, a verification set and a test set, and performing preprocessing; constructing a multi-task intensive prediction model; training the model by using a training set, and regularly evaluating the performance of the model in the training process by using a verification set; carrying out loss calculation and parameter optimization on the model in the training process; and performing final performance evaluation on the trained model by using the test set, and applying the trained model to a to-be-reasoned image to obtain an output result of the image under each dense prediction task. According to the method, a model comprising a shared feature extraction module, a task feature enhancement module and a task prediction module is constructed, and a mixed single-rank expert group is adopted to dynamically enhance task features in space and channel dimensions, so that a unified learning framework for performing multi-task intensive prediction on images is realized.