194 results about "Multiple experts" patented technology

A system and method for extending AI-enhanced decision platforms with deontic and normative reasoning capabilities that enhance adjustably autonomous decision-making through a novel integration of symbolic and neural approaches. The invention uses hierarchical and fuzzy deontic logic implementations alongside connectionist AI / ML to manage obligations, permissions, and prohibitions while maintaining observer awareness to achieve goals while incorporating knowledge across multiple expert domains. The system employs dynamic event and spatio-temporal knowledge graphs along with debate mechanisms, enabling high-assurance automated reasoning while preserving explainability through neuro-symbolic integration. In at least one embodiment, the invention operates through a federated distributed computational graph architecture that allows for arbitrary scaling while maintaining coherence, consistency and supporting compound workflows. The invention provides a framework for AI systems to make logically consistent, ethically-aware decisions by combining deontic reasoning with multi-agent coordination, token space communications and knowledge, including on intermediate results, enabling automated decision-making for a variety of applications.

A system and method for extending AI-enhanced decision platforms with deontic and normative reasoning capabilities that enhance adjustably autonomous decision-making through a novel integration of symbolic and neural approaches alongside quantum-inspired token management. The invention uses hierarchical and fuzzy deontic logic implementations and quantum-inspired state representations that combine complex amplitudes and phase information to manage obligations, permissions, and prohibitions while maintaining observer awareness to achieve complex goals while incorporating knowledge across multiple expert domains. The system employs dynamic event and spatio-temporal knowledge graphs along with debate mechanisms, enabling high-assurance automated reasoning while preserving explainability through neuro-symbolic integration and information-theoretic metrics. The platform is capable of operating through a federated distributed computational graph architecture that allows for arbitrary scaling while maintaining system coherence and logical consistency using quantum-inspired token operations and phase alignment transformations for optimizing information transfer between states.

The invention relates to a training method, a data processing method, electronic equipment and a computer readable storage medium, and relates to the technical field of computers. The training method comprises the steps that a training sample is input into a to-be-trained model, forward propagation is executed, output of the to-be-trained model is obtained, the to-be-trained model comprises a plurality of expert layers, and a specific activation value is not stored in the forward propagation process; determining the value of a loss function according to the output of the to-be-trained model; according to the value of the loss function and the loss function, back propagation is executed, in the back propagation process of each expert layer, a recalculation task of a specific activation value and a first communication task are executed in an overlapping mode, and for each expert layer, the first communication task comprises a communication task of data interaction among multiple devices corresponding to the expert layer.

The invention discloses a mixed expert model training optimization method based on matrix routing and Token allocation, and the method comprises the steps: firstly, in a pre-training process, an MoE model processes different input data through a plurality of expert sub-models, generates an expert index through a gating network, and generates a scheduling index and a combined index based on the expert index; secondly, based on the generation of the scheduling index and the combined index, entering a token scheduling and combined token stage, scheduling the token to each expert for expert processing, and combining the tokens processed by the experts by the combined token stage to generate a combined token; and finally, after pre-training of the MoE model is completed, fine tuning is carried out, and training optimization is completed. According to the method, on the premise that the fine adjustment precision of the model is kept unchanged, expert computing resources are fully utilized, and resource idling is effectively avoided.

The invention provides a request processing method and device based on a hybrid expert model, equipment and a medium, and relates to the technical field of artificial intelligence such as large language models, hybrid expert models, user portraits and knowledge maps. The method comprises the following steps: acquiring a target portrait of a user initiating a to-be-processed request; performing association expansion on the to-be-processed request according to the target portrait by using a preset knowledge graph to obtain a complementation demand including an expanded demand, and recording association relationships between different user entities and different behavior entities and between different preference entities in the knowledge graph; according to the intention actually expressed by the completion demand, determining a target vertical class to which the intention belongs; and sending the completion demand to a target expert sub-model corresponding to the target droop class in the current hybrid expert model for processing to obtain a processing result. According to the method, the accuracy of accurately positioning the target expert sub-model matched with the actual demand of the user in the multiple expert sub-models is improved, and then the accuracy of the request processing result is improved.

The embodiment of the invention discloses a collaborative reasoning method and device based on a multi-expert model and a medium, and relates to the technical field of collaborative reasoning, and the method comprises the steps: receiving a plurality of reasoning task requests, so as to obtain the task feature information of each reasoning task; based on the task feature information of each reasoning task and pre-acquired model registration information, performing task allocation to determine at least one allocation expert model corresponding to each reasoning task; predicting the model load of each distributed expert model, adjusting at least one distributed expert model corresponding to each reasoning task according to the predicted load rate, and determining at least one corresponding target expert model; and issuing each reasoning task to the corresponding target expert model, collecting single model reasoning data of each target expert model through an asynchronous message queue, and determining a collaborative reasoning result corresponding to each reasoning task through the single model reasoning data of each target expert model.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a signal processing method, device, equipment and medium based on mixed experts. And generating routing information based on the channel information, dynamically fusing expert output, generating a first task processing result by using the first task processing network, generating a second task processing result by using the second task processing network, jointly optimizing all network parameters, and generating a hybrid expert model for a to-be-processed signal. According to the method, the hybrid expert architecture aiming at the multi-channel characteristic is introduced, and the gating network is combined to dynamically allocate the characteristic processing path, so that the self-adaptive optimization aiming at different channel environments is realized, the accuracy of cross-channel identification and the resource utilization efficiency of the whole model are effectively improved, and the robustness of the cross-channel identification is improved. And high-precision, high-robustness and efficient signal identification can be realized in a complex and changeable environment.

The invention provides a federated learning method and system based on a hybrid expert model, and relates to the technical field of machine learning, and the method comprises the steps: carrying out the initialization of a local gating network of a current client through employing a global gating network, obtaining an initial local gating network # imgabs0 #, carrying out the feature extraction of a local data set, and obtaining a local data set; k target expert models Sk with the highest adaptation degree with the data features are screened out from the multiple expert models or a required target expert model # imgabs1 # is downloaded from a server side by using a Top-K-based adaptive expert selection mechanism to ensure that all the selected expert models can be used locally; and the client uses local data to perform forward calculation on the target expert model, obtains training feedback parameters of the target expert based on a prediction result and a real label, and optimizes the target expert model through a back propagation algorithm # imgabs2, and the client performs weighted summation on calculation results of the target expert models to obtain a comprehensive prediction result. And the generalization ability and performance of the model are improved.

The invention discloses a data processing method and system for a low-energy-consumption large language model based on a momentum mechanism and multiple types of experts, and belongs to the technical field of large language models, and the method comprises the steps: obtaining a hybrid expert model which comprises a gating network, a combination network and a plurality of expert networks; obtaining target data based on a lazy loading mechanism; obtaining fitness scores of the target data corresponding to the expert networks through the gating network, and obtaining the expert network meeting the sorting requirement as a target network; and obtaining output data of the target data in the target network, and carrying out weighted summation on the multiple output data through the combined network. According to the method, the multiple expert networks are arranged in the hybrid expert model, the fitness ranking is performed on the fitness scores of the target data, and the output of the expert network meeting the ranking requirement is selected to perform weighted summation so as to obtain the final output, so that the floating-point operation and calculation memory overhead are reduced, the calculation efficiency is improved, and the memory resource waste is reduced.

The embodiment of the invention provides a large language model training method and device, computer equipment, a medium and a product, and relates to the technical field of model training. According to the method, an initial MOE model is pre-trained through text data to obtain an MOE pre-training model, and then multi-task training is performed on the MOE pre-training model through a second training set, so that an MOE dialogue model obtained through training can generate an output result related to input data according to the input data, and various different tasks are completed. And finally, performing user preference guidance on the MOE dialogue model to obtain an MOE reward model so as to evaluate an output result of the MOE dialogue model, thereby optimizing model parameters of the MOE dialogue model. In the embodiment of the invention, the MOE model is used for replacing a traditional dense model, different functions of the model are realized through multiple expert sub-models in the MOE model, and multiple functions are realized by using fewer reasoning video memories.

The invention belongs to the field of natural language processing, and particularly discloses a task processing method and system based on multiple expert layers, a terminal and a medium, a multi-layer expert network is constructed, and the multi-layer expert network comprises a basic expert layer used for processing general field tasks, a field expert layer used for processing a plurality of specific field tasks and a task expert layer used for processing a plurality of task types; receiving an input task, and analyzing the input task to obtain task characteristics of the input task, including a task field, a task type and task complexity; formulating an expert selection strategy according to the task characteristics; and according to the expert selection strategy, inputting the input task into the target expert layer for processing to obtain a processing result. According to the method, the multi-layer expert network is constructed, the corresponding expert layer is selected according to the task characteristics, adaptation to multiple fields and multiple tasks is achieved, the flexibility and pertinence of the model are improved, and the task processing efficiency is improved.

The invention discloses an expert model training method and device, a storage medium and electronic equipment, and relates to the technical field of computers, and the method comprises the steps: determining a target expert network group from a plurality of expert network groups included in an initial expert model according to an estimated resource occupation condition of input data; taking a plurality of expert networks in the target expert network group as a current expert network in sequence, and determining an individual difference evaluation coefficient matched with the current expert network; determining at least one target expert network from the plurality of expert networks according to the individual difference evaluation coefficients of the plurality of expert networks in the target expert network group; and processing the input data according to the at least one target expert network, and adjusting the model parameters in the initial expert model according to the load state description information corresponding to the at least one target expert network, thereby solving the technical problem of inaccurate selection of the expert network in the prior art.

A future workload may be predicted for a database system using analysis of queries submitted for execution. A feature vector for a query may be determined according to a query plan for the query. If the feature vector has not been previously seen, or has not been sufficiently seen, by the database system, a machine learning inference may be used to predict performance characteristics of the query, the machine learning system trained using previous feature vectors and performance characteristics of executed queries. If the feature vector has been sufficiently seen previously by the database system, a history of performance characteristics of previous queries with similar or the same feature vector may be used to predict the performance characteristics. The predictions may then be used to configure or reconfigure processing cluster(s) of the database system.

The invention discloses an expert parallel computing method for a hybrid expert model and a computer program product, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring a storage address of each input data block in a global sequence based on the input data block processed by each expert sub-network and an expert identification sequence corresponding to a plurality of expert sub-networks; based on the occurrence frequency of each expert sub-network in the expert identification sequence and the storage address corresponding to each input data block, obtaining input fragments processed by all expert sub-networks deployed on each computing card; distributing a target video memory space for each computing card based on the input fragment corresponding to each computing card; controlling each expert sub-network deployed on each computing card to read the input data block needing to be processed from the target video memory space corresponding to each computing card, and calculating the input data block needing to be processed to obtain sub-result data output by each computing card; and performing fusion processing on the at least one piece of sub-result data to obtain target result data output by the hybrid expert model.

A computer-implemented system and computer instructions stored on non-transitory computer readable medium for bilingual medical inquiry in both Arabic and English, including multiple-choice question answering, open-ended question answering, and multi-turn question answering. The system and instructions use a mixture of experts large language model (MOE LLM) having a router network connected to multiple expert networks. The MOE LLM is trained with medical domain data and is used to receive the input bilingual text in a format for a medical inquiry, and output text in a format of a response to the medical inquiry, in sequence. The system and instructions incorporate an English-to-Arabic translation pipeline having a language translation model to generate Arabic language medical instruction sets from English language medical instructions, for large scale use in Arabic and English medical inquiry.

The invention provides a medical image automatic segmentation method and device based on a variational hybrid expert model, and relates to the technical field of medical image processing. The method comprises the following steps: acquiring a three-dimensional medical image, and inputting the three-dimensional medical image into a U-Net encoder for multi-level convolution, down-sampling feature extraction and adaptive dictionary enhancement operation to generate each level of enhanced feature map; inputting the enhanced feature map of the last level into a variational hybrid expert module to perform shared feature extraction, and outputting and fusing with dynamic routing selection results of a plurality of expert models to generate fused features; inputting the fused features and the enhanced feature map of the last level into a U-Net decoder for up-sampling operation, and splicing the output of up-sampling and the enhanced feature map of the current level step by step to recover the spatial resolution of the image to obtain decoded fused features; and inputting the decoding fusion feature into a single-channel convolutional layer, and outputting a segmentation result. According to the method, the accuracy and robustness of medical image segmentation and the adaptive capacity to the multi-modal medical image are effectively improved.

The invention discloses a multi-type context-aware dialogue recommendation method based on a hybrid expert model. The aim of the invention is fulfilled by a hybrid expert framework comprising a plurality of expert module coordination systems. The knowledge graph expert model is responsible for extracting and coding structured entity relationship information, the dialogue expert model is used for deeply mining key contents and user intentions in dialogues, and the comment experts are used for analyzing user emotion and preference information in commodity comments. In this way, the system can make full use of semantic information of the structured data and the unstructured data, and the problem of semantic gaps caused by data heterogeneity in the prior art is solved. On the basis, a coordination system is introduced to serve as a core module for coordination and integration. The coordination system can dynamically allocate the output of the expert module according to the weights and contributions of different context information to generate a final recommendation result, so that efficient fusion of multiple types of context information is realized, the problems of semantic alignment and inconsistency between different types of data are solved, and the performance of the dialogue recommendation method is improved.

The invention discloses a pulmonary nodule risk prediction method, system and device based on a large model, and a medium, belongs to the technical field of artificial intelligence, and aims to solve the technical problems of insufficient generalization ability of a network model and poor pulmonary nodule multi-task prediction effect in the prior art. The method comprises the following steps: acquiring sample data of multiple modalities including medical images, gene detection and demographic data; single-mode features are obtained through an exclusive feature extraction module, and then the single-mode features are input into a plurality of expert networks to obtain single-mode experts; obtaining a multi-modal feature after the single-modal feature passes through a cross-modal attention module, and inputting the multi-modal feature into a plurality of expert networks to obtain a plurality of multi-modal experts; the three task MoE modules are all provided with task gating networks in a matched mode, and the task MoE modules select the most appropriate expert set according to the calculated probability that each expert is activated; acquiring to-be-predicted data, inputting the to-be-predicted data into the selected expert set, and outputting malignant, growth and wettability risk prediction results of the pulmonary nodules by the task model.

The present invention belongs to the field of image classification and designs a long-tail image data classification method based on mixed samples. The present invention proposes a new solution to the long-tail training set problem encountered in image classification research. It aims to use three experts with specialized knowledge to jointly assist the algorithm in making the final decision, avoiding problems such as excessive deviation of the model classifier weight caused by a single model. The present invention is suitable for business scenarios of image classification with long-tail data distribution. By designing multiple experts with specific field knowledge, the classification performance of the model for all frequency distribution types is improved without losing the accuracy of the head class classification. It provides a solution for the actual engineering application of image classification when the data has a long-tail distribution, alleviates problems such as data collection difficulties, improves the overfitting of the algorithm model to the head class data, and improves the learning ability of the tail class data.

The invention discloses an automatic diagnosis algorithm for interpretability-oriented adaptive multi-agent cooperation. The method comprises the following steps: firstly, constructing a bidirectional probability network of symptoms and diseases based on a data set, calculating a disease weight through a pre-diagnosis module, and deciding to enter a Doctor Agent module or a Med-Team Agent module; the Doctor Agent simulates single expert diagnosis and judges whether a patient can be diagnosed or not or continues to inquire related symptoms to assist diagnosis in combination with a medical knowledge base and a bidirectional probability network. The Med-Team Agent is used for simulating consultation of multiple experts, and comprises a Manager Agent and a plurality of Disease Agents which respectively represent an attending physician and a plurality of disease experts. And the Manager Agent dynamically manages a plurality of Disease Agents, selects an appropriate expert to generate related symptoms according to the symptom information of the patient, and finally completes diagnosis. The method is suitable for medical automatic diagnosis tasks, through combination of multi-agent cooperation and bidirectional probability network prompting, the interpretability of diagnosis is improved, and efficient inquiry and accurate diagnosis are achieved.

The embodiment of the invention relates to the field of computing power scheduling, and discloses an intelligent inference system and method, equipment and a medium. The system includes a base model for performing an inference process of shared parameters; a plurality of adapter models, each of which is used for performing an inference process of an inference task; the base model and the adapter model are separately and independently deployed, and parallel inference of the multiple adapter models on one base model is supported; and the intelligent router is used for receiving and analyzing the inference task, selecting the corresponding adapter model according to a preset rule, and loading the adapter model to the corresponding inference node. By introducing the adapter design of the LoRA model and the intelligent router mechanism, the efficient deployment and inference of the multi-expert model in the low video memory environment are realized, the use of the video memory can be obviously reduced, and the inference response speed is improved at the same time. The technical problem of low expert model inference efficiency can be at least solved.

The invention relates to the technical field of artificial intelligence, and particularly provides a data processing method of a large language model based on an improved hybrid expert architecture. The method comprises the following steps: processing input query information through a semantic understanding module to obtain an understanding result; based on the understanding result and a pre-stored knowledge graph, the query information is sent to a plurality of corresponding expert models, corresponding output results are obtained by the expert models, the plurality of expert models are arranged in multiple layers based on an improved hybrid expert architecture, and the expert models of different layers are obtained through division according to different expert dimensions; and carrying out fusion processing on the output results of the plurality of expert models to obtain an answer result corresponding to the query information. The problems that a large language model of a hybrid expert architecture in the related technology is low in accuracy and poor in problem solving capacity in the professional subdivision field are solved.

The invention relates to the technical field of industrial process prediction, in particular to a tundish erosion prediction method based on a hierarchical hybrid expert framework. The method comprises the following steps: collecting time sequence physical field data of the tundish, and screening features to construct a unified feature space; shunting the feature space to obtain a time sequence feature and a statistical aggregation feature; a statistical aggregation feature training classifier is utilized to generate a calibration posterior probability, and a gating network is constructed; generating an initial mode subset based on a posterior probability and training a corresponding expert model; the confidence of data to be measured is obtained by the gating network, and a single expert model is selected for prediction or multiple expert models are fused through a self-adaptive strategy for weighted prediction according to whether the confidence exceeds a threshold value or not; and finally, reconstructing the predicted value into an erosion thickness absolute value through inverse transformation. According to the method, the accuracy and adaptability of tundish erosion prediction are effectively improved.

The invention provides a laser powder bed melting method for preparing a high-toughness TC4 titanium alloy based on machine learning. The method comprises the steps that performance data of the TC4 titanium alloy prepared through different SLM process parameters are collected; constructing a hybrid expert (MoE) machine learning prediction model framework; classifying the TC4 data by adopting a K-means clustering algorithm, and taking a clustering center as an initial weight of an input layer and a first layer of the gating network; presetting a TC4 performance prediction model through the gating network and the plurality of expert networks after clustering information initialization; performing small-range iteration through large-step network search in combination with a Bayesian optimization method so as to determine an optimal model structure and complete model training; the trained performance prediction model and a genetic algorithm are used for carrying out backstepping to obtain SLM printing parameters of the high-toughness TC4 alloy; and SLM forming is carried out, and the high-toughness TC4 titanium alloy is prepared. According to the method, a calculation framework which is high in precision and suitable for exploring a high-unknown characteristic space is constructed, and far-reaching influences on titanium alloy additive manufacturing and process optimization of other alloy systems are achieved.

A 3D in-memory computing accelerator system (100) and method for efficiently inferring a hybrid expert (MoE) neural network model. The system includes a plurality of in-memory computing cores (102), each memory core including a plurality of layers of in-memory computing units (106). One or more layers of the in-memory computing unit correspond to expert sub-models of the MoE model. One or more expert sub-models for activation propagation are selected according to a function-based routing (115) based on which a corresponding expert layer is activated. In one embodiment, the function is a hash-based layer selection function for dynamic routing of input and output activation. In an embodiment, the function is applied to select a single expert or multiple experts for single-level activation using input data based or using MoE based on layer activation. Further, the system is configured as a multi-model system with a single expert model selection or a multi-model system with multiple expert selections.

The invention relates to an airport core risk whole-process management method based on a dual-prevention working mechanism, and the method comprises the following main steps: a data collection step: collecting multi-source data related to the operation of an airport in real time through a plurality of sensors and systems; a data fusion step of replacing a multi-head attention mechanism in a traditional Transform model with a plurality of expert large models, including feature extraction and analysis, information aggregation and data synchronization and standardization; the method comprises the steps of generating adversarial data, generating diversified virtual risk scenes and adversarial samples by utilizing a plurality of expert large models, automatically constructing a virtual environment, automatically constructing a high-fidelity virtual simulation environment based on the large models, and performing risk assessment and decision making. Accurate risk assessment and decision support are carried out by using a self-adaptive risk scoring system and an intelligent decision support mechanism; the comprehensive, accurate and intelligent management of the airport core risk is realized, and the safety and the emergency response capability of airport operation are obviously improved.

The invention discloses a mechanical equipment residual life prediction method based on multiple expert models, and belongs to the technical field of mechanical equipment. Comprising the steps of receiving an original mechanical vibration signal, performing preprocessing, constructing a residual life prediction model of the mechanical equipment, training the residual life prediction model of the mechanical equipment based on a preset data set, inputting a mechanical vibration signal to be detected into the trained residual life prediction model of the mechanical equipment, and performing fault category prediction of the mechanical equipment. According to the method, the accuracy, generalization and practicability of residual life prediction of mechanical equipment are remarkably improved, and through multi-scale time-frequency domain feature fusion and a double attention mechanism, the identification capability and noise immunity of the model to a complex fault mode are enhanced; the domain adaptive routing network realizes cross-device efficient knowledge migration, greatly improves the prediction performance on unseen devices, has high precision and high real-time performance, and provides reliable technical support for industrial predictive maintenance.

The invention relates to the technical field of software vulnerability detection, and discloses a software vulnerability detection method and device based on a multi-modal hybrid expert model. The method comprises the following steps: constructing and training a multi-modal hybrid expert network to obtain a multi-modal hybrid expert model; converting the target software code into different modal representations and inputting the different modal representations into a multi-modal hybrid expert model to obtain a vulnerability detection result of the target software vulnerability; wherein the multi-mode hybrid expert network comprises a vulnerability detection module and an expert gating routing module; the vulnerability detection module comprises a plurality of expert layers, and different expert layers are used for detecting vulnerabilities in different modes; the expert gating routing module comprises an input routing strategy and an expert decision output fusion strategy, the input routing strategy distributes the input of the multi-mode hybrid expert model to different expert layers of the vulnerability detection module based on different modes, and the expert decision output fusion strategy carries out weight distribution on the output of the different expert layers of the vulnerability detection module; and obtaining a vulnerability detection result.

The invention discloses a code generation method and device and computing equipment, and the method comprises the steps: processing a received input content through a universal coding model, and obtaining a context representation; the universal coding model is obtained by training the training sample data of multiple programming languages; selecting a target expert adapter from a plurality of expert adapters according to the context representation; different expert adapters are obtained by training the training sample data of different programming languages; processing the context representation by using a target expert adapter to obtain language detailed features; and fusing the context representation and the language detailed features, and generating a target code according to features obtained through fusion. By means of the mode, the universal features and the unique features of the specific programming language are fused for code generation, the correctness of code generation can be improved, part of expert adapters are lightly activated instead of all expert adapters, the code generation speed can be increased, and calculation consumption can also be reduced.

The invention discloses a routable distributed fingerprint embedding and migrating method and device based on hybrid experts, and belongs to the technical field of large language model intellectual property protection. The basic model is transformed based on the hybrid expert system architecture, the basic model is divided into a plurality of expert sub-modules and a gate control module, at least one expert sub-module is designated as a fingerprint expert module, and the gate control module is used for routing the trigger sample to the corresponding fingerprint expert module; the fingerprint expert module is finely adjusted based on the fingerprint data set, and model parameters embedded with fingerprint features are stored in an independent low-rank adapter; combining the fine-tuned fingerprint expert module and the low-rank adapter thereof with the expert sub-module of the corresponding level in the downstream model to realize fingerprint migration; and inputting the trigger sample into the downstream model after fingerprint migration for copyright verification and evaluation. According to the method, the model performance, the security and the verifiability are considered, and the adaptability of the fingerprint technology in an industrial large model deployment scene is improved.