100 results about "Neuronal activation" patented technology

The invention discloses a federated learning backdoor defense method based on pruning and fine tuning in the technical field of artificial intelligence and network security, the method realizes defense through two core mechanisms of dynamic pruning and gradient constraint fine tuning, and the method comprises the following steps: firstly, calculating a sensitivity score based on a neuron activation frequency and a weight outlier degree; dynamically identifying and cutting redundant neurons utilized by a backdoor, and blocking an abnormal activation path; secondly, gradient direction consistency detection and amplitude constraint are introduced in the fine tuning stage, and a malicious client is inhibited from reconstructing a back door through an abnormal gradient; the server continuously purifies model parameters and enhances robustness by cyclically executing pruning, fine tuning and aggregation operations; the method does not need to depend on an extra clean data set, strictly follows a federated learning privacy protection principle, reduces communication overhead through lightweight pruning, maintains main task performance in combination with gradient constraint, is suitable for a federated learning scene in which edge equipment participates, and effectively balances a defense effect and model stability.

The invention relates to the technical field of artificial intelligence model optimization, and discloses a multi-granularity dynamic pruning method and system for a generative AI model. The system comprises a model state acquisition module which acquires output of a middle layer in real time through a probe, and constructs a feature information set of neuron activation distribution, weight matrix norm and connection topology; the sparseness evaluation module outputs a sparseness risk value based on the feature information set, the initial sparseness parameter and the real-time computing resource state; a pruning planning collaborative analysis screening module generates a multi-granularity pruning scheme when the risk value exceeds a threshold value, calculates a collaborative interference value of precision recovery operation and screens an optimal scheme; and the pruning strategy execution feedback module updates model parameters, generates logs and feeds the logs back to the management terminal. The system realizes adaptive structure optimization and resource scheduling of the generative model in the reasoning process.

The invention relates to the technical field of artificial intelligence interpretability, and discloses a neuron correction-based large language model cue word attack prevention method, which comprises the following steps of: constructing an attack cue word data set, and rewriting attack cue words to obtain a contrast cue word data set consisting of contrast cue word data; constructing a probe, obtaining output values of all neurons in the large language model during reasoning, and further obtaining a neuron activation state of the large language model; inputting the attack cue word and the corresponding contrast cue word into a large language model, and summarizing the obtained neuron activation state table to obtain an additional activation neuron corresponding to the attack cue word; and inputting cue word attack data into the large language model, and forcibly zeroing the output of one or more neurons in the additional activation neurons. By correcting the key neurons, the model performance and the prevention effect are balanced, and the security of the large language model is improved.

The invention discloses a neural network training method based on a back propagation algorithm, and relates to the technical field of network training, and the method comprises the steps: obtaining the neuron output data of each neural network layer through a forward propagation process, and constructing the output distribution data reflecting the activation features of each layer of neuron; obtaining error data based on a back propagation process, establishing error term distribution of each neural network layer, and constructing an entropy change estimation factor for measuring network information state change; performing disturbance control on the error data according to the value trend of the entropy change estimation factor, and dynamically adjusting the parameter updating amplitude and direction of the neural network by the processed error data and the output distribution data; and weight iteration of the neural network is guided through the parameter updating behavior. According to the method, error disturbance control and dynamic parameter adjustment are combined, a whole set of mechanism of information perception-disturbance control-dynamic optimization is formed, and dynamic perception and accurate guidance of information state evolution in the neural network training process are achieved.

The invention relates to the technical field of calculation, and discloses a continuous time digital Isin model hardware solver based on combinational logic, which comprises a plurality of neurons; the data initialization module is used for carrying out initialization assignment on corresponding coefficients of the hardware solver; after the control signal is used for carrying out initialization assignment on neurons, iterative calculation is started; the neuron updating module is used for carrying out matrix operation on an interaction relation weight matrix formed by the initialized weight coefficients and initialized neurons, and adding the matrix operation and bias to obtain the cumulative sum of each neuron in the current neuron state; the neuron activation module is used for adding nonlinear factors for iterative calculation; the high precision of a digital circuit is utilized, but the continuity of a physical system is also utilized at the same time, so that acceleration hardware which is lower in energy consumption and faster in digit solving is realized.

The invention relates to the technical field of geographic information systems, and discloses a geographic information data processing method and system.The geographic information data processing method comprises the steps that geographic data are packaged into self-activation data nodes; monitoring a network state in real time through a network environment perception function; constructing an activation threshold function based on a neuron activation principle to determine data transmission; carrying out multi-scale analysis on the data stream by applying wavelet transform to realize self-adaptive compression; predicting hotspot distribution dynamic scheduling resources through space-time analysis; an edge-fog-cloud three-level cooperative processing architecture is constructed to realize distributed processing; according to the method, the technical problem of geographic information data transmission and processing in an unstable network environment is solved, the overall performance and the adaptive capacity of the system are improved through a multi-level adaptive mechanism, and the method has wide application value.

The invention discloses a method and system for improving image classification performance by using a spiking neural network, and the method comprises the steps: focusing on a neuron activation mode and synaptic plasticity in an unsupervised handwritten numeral recognition research process based on a triple pulse time-dependent plasticity learning rule, and after the fluctuation of accuracy is observed, carrying out the recognition of an unsupervised handwritten numeral recognition model, and carrying out the recognition of the unsupervised handwritten numeral recognition model. According to the method, key influence factors such as weight initialization, STDP randomness and hyper-parameter sensitivity are deeply analyzed; according to the invention, a common SNN structure is optimized, a pulse Inception module is realized, and the SNN is trained through a triple STDP; according to the method, the simulated annealing algorithm and the STDP are innovatively combined to be used for neuron parameter optimization, an optimal hyper-parameter combination is established through SA, after STDP training, the performance is remarkably improved under the condition of a limited data set, and particularly, the high efficiency of one-time perceptual classification of hundreds of digits is shown.

The invention provides a visual motion function test system and method based on full-view surround synchronous visual stimulation, and is applied to the technical field of medical data processing. Visual stimulation parameters and rotating bar motion mode parameters are combined, the rolling direction of a bar and the rotating direction of a rotating bar are synchronously matched, a synchronous presentation scene of visual stimulation and motion tasks is established based on a four-screen linkage control technology and a rotating bar-screen adaptive design, and full-view stimulation constraint conditions are generated; processing is carried out based on visual stimulation parameters, rotating rod reference data, synchronous control logic and dynamic parameter combination constraint conditions, and rod time, rotating speed during falling and motion trail data are collected; processing the collected kinematics data, and combining functional parameters including visual state grouping and brain region neuron activation counting to generate motion balance ability and brain function associated data; and processing by combining normality test and an inter-group statistical method to generate a vision-motion function evaluation result.

A transcranial magnetic stimulation (TMS) treatment system is provided. The system includes a sensor device that senses EEG signals from a subject through one or more leads and a server device configured to receive EEG data corresponding to the subject. The server includes an analysis module configured to process the EEG data and determine a personalized resonant brain frequency and a minimum neuronal activation threshold of the subject based at least in part on EEG data corresponding to one or more leads of the sensor device. The analysis module is also configured to determine a TMS treatment protocol where the treatment protocol includes at least a frequency based on the personalized resonant brain frequency and an amplitude based on the minimum neuronal activation threshold. The system also includes a treatment device configured to deliver a TMS treatment to the subject based on the TMS treatment protocol received from the server.

The invention discloses a method and a device for decoupling a slow-thinking area and a quick-response area in a large language model. The method comprises the following steps: constructing a slow thinking prompt and a fast response prompt, respectively inputting the slow thinking prompt and the fast response prompt into a large language model, and recording a neuron activation value of each linear layer of the large language model; calculating an activation value difference vector based on the neuron activation values corresponding to the slow thinking prompt and the fast response prompt, and obtaining a prompt difference vector based on the activation value difference vector; iteratively processing a plurality of reasoning problems, obtaining positive value position intersection of prompt difference vectors in all reasoning problems, and identifying a key neuron set related to slow thinking; and constructing a gradient mask matrix based on the position information of the key neurons in the key neuron set, and performing fine tuning on the key neurons by adopting a mask mode based on the gradient mask matrix to obtain an optimized large language model. The technical problem that an existing model is poor in reasoning ability is solved.

The invention relates to a neural network fuzzy test seed selection and anti-disturbance mutation method. The method comprises four stages of uncertainty evaluation, redundancy elimination, mixed mutation and dynamic sorting. According to the invention, through the seed selection model of coverage-uncertainty joint optimization and the gradient-adversarial hybrid mutation strategy, the efficiency and defect discovery capability of the deep neural network fuzzy test are significantly improved; the problems that in an existing coverage rate guide fuzzy test, seed selection lacks directivity, redundant samples are too many, and a mutation strategy is disjointed with a coverage target are solved. Compared with an existing most advanced method, the method has the advantages that the coverage rates of five indexes including neuron coverage, k-multi-segment neuron coverage, neuron boundary coverage, strong neuron activation coverage and top layer k neuron coverage are averagely improved by 4.56%-18.53%, SNAC is improved by 22.95% on the LeNet1 model, and more decision boundary violation behaviors are revealed.

The present invention proposes a fast memory encoding method based on a multi-synaptic plasticity pulse neural network, including: step one: converting external stimuli into an input pulse sequence based on a hierarchical encoding strategy; step two: after the pulse neural network receives the input pulse, the membrane potential of the output layer neurons is updated based on the improved SRM model; step three: using the supervised group Tempotron to update the synaptic weights between the input and the output layer, and activating the output layer neurons to memorize the input; step four: after the output layer neurons are activated, using unsupervised STDP to update the synaptic weights between the activated neurons in the layer to form an enhanced recurrent subnetwork to store memory; step five: while executing step four, using unsupervised inhibition of synaptic plasticity to update the synaptic weights between the inhibition layer and the output layer, and inhibiting feedback to ensure the separation of the firing time of neural groups with different inputs. The present invention also proposes a fast memory encoding device based on a multi-synaptic plasticity pulse neural network. The present invention effectively improves the encoding speed and stability of memory.

The present invention relates to a communication signal recognition method and system based on an adaptive feedforward neural network, belonging to the technical field of signal recognition, and solves the problems of inaccurate recognition of existing small-sample signals and difficult adjustment of model parameters. It includes randomly generating the activation states of each hidden-layer neuron of binary type, the input weights of each hidden layer of real number type, and the hidden-layer bias based on a single-hidden-layer feedforward neural network, and forming multiple initial network individuals in a hybrid coding form as the parental population; using the root mean square error and network complexity as the objective function, and based on the preprocessed signal sample set, an improved NSGA-III algorithm is used to obtain the Pareto optimal population; according to the root mean square error, multiple network individuals are obtained from the Pareto optimal population as base classifiers, and the preprocessed communication signals collected in real time are input into the base classifiers for ensemble learning. After weighted summation of the output results, the category corresponding to the maximum value is taken as the communication signal recognition result. Accurate signal recognition is achieved.

The invention discloses a hardware acceleration system and method supporting a pulse-gated recurrent neural network, and relates to the technical field of neural network hardware acceleration, and the system comprises a PC terminal and an accelerator. The accelerator comprises a UART (Universal Asynchronous Receiver / Transmitter) module, a data composer, a storage module, a controller module, a calculation module and a gating calculation core, the PC terminal is connected with the UART module; the gating calculation core is connected with the data composer; and the controller module and the storage module are respectively connected with the UART module, the data composer, the calculation module and the gating calculation core. Based on the above scheme, different coding modes can be selected at a PC terminal to adapt to different data types, and a pulse neuron activation function and a logic gate are utilized to perform calculation heterogeneous design in a calculation module and a gating calculation core, so that a high-performance configurable hardware acceleration calculation architecture for the pulse gating recurrent neural network is realized.

The present invention discloses an interpretation method for convolutional neural network image recognition classification decision, firstly, an image perturbed by a feature map is input into a neural network model, and a neuron activation value is obtained through a classification output layer; then, the neuron activation value is divided by a penalty factor corresponding to the feature map, and then input into a softmax layer to obtain a score of a target class as a weight of the feature map, and finally, an interpretation map is obtained according to the weights of a plurality of the feature maps. The present invention realizes the difference between the average values ​​of feature maps as required by a parameter, adjusts the importance of the softmax function to enlarge / reduce features, greatly reduces the amount of calculation, and the method can generate a heat map and an interpretation map according to the standard desired by the user, has very good applicability and very high interpretation credibility, and in addition, the method can be used for any convolutional neural network without restrictions on the model structure.

The invention relates to a technique for improving confidence estimates associated with neural networks. The technique involves computing neuron activation statistics during training, evaluating neuron activations during inferencing and determining how the activations compare with the previously computed statistics (e.g. whether prediction activations are within the bounds of the training activation statistics). The comparison may be used to compute a confidence value for the neural network.

There are provided systems and methods for data privacy protection and removal for artificial intelligence model training and deployment. An online transaction processor or other service provider may provide computing services and platforms to entities, which may include use of machine learning (ML) models including large language models (LLMs). To comply with data privacy protections and copyright enforcement, a system may provide unlearning of content from ML models. The system may receive a request to unlearn a content and, after verifying the request is valid, identify the content used for during training of or inferencing by an ML model. The system may then map the content to concepts and correlate those concepts with ML model outputs using projections in a vector space. Based on the mapped concepts and outputs, neuron activation of the ML model may be analyzed to identify a negation vector and perform selective parameter dampening.

A transcranial magnetic stimulation (TMS) treatment system is provided. The system includes a sensor device that senses EEG signals from a subject through one or more leads and a server device configured to receive EEG data corresponding to the subject. The server includes an analysis module configured to process the EEG data and determine a personalized resonant brain frequency and a minimum neuronal activation threshold of the subject based at least in part on EEG data corresponding to one or more leads of the sensor device. The analysis module is also configured to determine a TMS treatment protocol where the treatment protocol includes at least a frequency based on the personalized resonant brain frequency and an amplitude based on the minimum neuronal activation threshold. The system also includes a treatment device configured to deliver a TMS treatment to the subject based on the TMS treatment protocol received from the server.

The embodiment of the invention discloses a large model output result detection method, and the method comprises the steps: determining the activation state of neurons of a to-be-detected large model LLM in a reasoning process based on a training sample marked with whether the output result is correct or not, and obtaining the relation between whether the output result is correct or not and the activation state of the neurons, the method comprises the following steps: training a detection classifier by using a neural network to obtain an activation state of a neuron when a to-be-detected LLM executes task input data, and detecting the correctness of an LLM output result through the detection classifier, namely whether illusion appears or not. According to the method, whether the illusion occurs in the LLM or not is identified by associating the activation state of the LLM neurons with the illusion occurring in the LLM, and whether the illusion occurs in the LLM or not is identified from the bottom layer of the model without paying attention to the LLM reasoning process and the output result.

Systems, methods, and computer-readable media for (i) accelerating the inference speed of deep neural networks (DNNs) and (ii) compressing vector representations generated by DNNs from various input data (such as images, audio, video, and text). The method embodiments take a neural network architecture and a task-related loss function as input, measure the performance of the neural network on a training data set, and output a deep neural network with sparse neuron activations. The process of the present invention enhances the existing training objective function of the DNN using a regularization term, promotes sparse activations of neurons, and compresses the DNN by using various algorithms to solve optimization problems. The present disclosure also shows how to exploit sparsification of activations during DNN inference, thereby proportionally reducing the number of arithmetic operations, and how to use the sparse representations generated by the DNN to build an efficient search engine.

The present invention belongs to the field of integrated circuits and relates to an efficient CNN multiply-accumulate unit based on logarithmic data. The present invention includes a linear domain to logarithmic domain conversion module, a logarithmic domain addition module, and an accumulation module. In the efficient CNN multiply-accumulate unit based on logarithmic data, the multiplication operation of multiplying the weight W and the neuron activation value X is converted into an addition calculation in the logarithmic domain. First, the neuron activation value X and the weight W in the linear domain are converted into the logarithmic domain, and then the adder is used to add the already logarithmized neuron activation value X and the weight W. Finally, the output data of the logarithmic domain addition module is shifted and accumulated to obtain the result of multiplying and accumulating the neuron activation value X and the weight W in the linear domain. The present invention can reduce the complexity of hardware design, hardware power consumption and area, as well as the maximum combinational logic delay, thereby increasing the maximum clock frequency of the system. It can process multiple groups of data in a pipeline form to achieve an improvement in the operation rate.

The invention relates to the technical field of capacity collaboration, and discloses a capacity collaborative optimization method and system based on a knowledge graph, and the method comprises the steps: constructing a capacity collaborative decision model based on the knowledge graph; designing a distributed homomorphic neural computing framework; developing a pulse neural network encoder; realizing a pulse neural network inference mechanism in a homomorphic encryption domain; designing a neuron activation mechanism based on a threshold value; constructing an inter-enterprise security communication protocol; developing a calculation result verification mechanism based on zero knowledge proof; a result decryption mechanism for realizing separation of rights and responsibilities; designing a special neuromorphic hardware acceleration chip; constructing a feedback updating mechanism of the decision result and the knowledge graph; according to the method, an enterprise can participate in collaborative decision-making on the premise of not leaking sensitive data through a secure computing technology, and meanwhile, the energy consumption and delay of complex computation are reduced by utilizing a neuromorphic computing architecture.

The invention discloses a road target detection method and device based on multi-sensor fusion, and belongs to the technical field of target detection, and the method comprises the steps: collecting the multi-modal sensor data of a target road and the position information of each sensor; based on the position information of each sensor and the multi-modal sensor data, analyzing a space-data communication relationship among the sensors, and constructing a sensing adjacency matrix; extracting spatio-temporal characteristics of the multi-modal sensor data based on the sensing adjacency matrix, generating a spatio-temporal characteristic set, and obtaining a neuron activation value of each calculation path in real time; performing redundancy analysis on each calculation path according to the neuron activation value of each calculation path, and screening out a reserved path; performing redundancy removal processing on the space-time feature set based on each reserved path to obtain a sparse space-time feature set; and performing data fusion on the sparse space-time feature set to obtain a target detection result of the target road. By implementing the method and the device, the problem of low target detection accuracy in the prior art can be solved.

The present invention relates to the field of neuronal computing, and in particular to a neuromorphic computing device. The device includes a neuronal state memory, a neuronal interconnection memory, a controller, a physical neuron, an external neuronal excitation input unit, and an external neuronal activation output unit. The biological state of all neurons in the neural network system is stored in the neuronal state memory, and the interconnection state of all neurons in the neural network system is stored in the neuronal interconnection memory. The device works periodically at a time interval of Δt, and completes the update of all neuronal states in the neuronal state memory within the Δt time: the controller reads the neuronal interconnection state, sequentially configures the physical neurons as corresponding virtual neurons, and the physical neurons actively obtain data from the state memory according to the configuration information, completes the corresponding calculation, and finally updates the state of the corresponding neuron. This device can realize flexible, universal and arbitrary scale neural network calculations.

The invention discloses an optimal path planning method of a pulse coupling neural network based on double constraints. The method comprises the following steps: mapping a path planning environment to a DC-PCNN network; a DC-PCNN neural network model is constructed, and all neurons are initialized; activating a target neuron, and recording the current neuron as a father node; calculating an exponential decay function of the torque deviation, and multiplying the calculated value as a penalty factor by an update item of the internal activity item; calculating a gravitational function value of the flow field constraint, and using the calculated value to update a current neuron dynamic threshold value; comparing the internal activity item with a dynamic threshold; when gt; if yes, activating the neuron, and recording the current neuron as a father node; repeating the steps S4-S6 until the initial neuron is activated; and backtracking all activated nodes, and planning an optimal path. According to the method, the search efficiency is remarkably improved while the path optimality is ensured.

The invention discloses oligopeptide and application thereof in preparation of a medicine for treating and / or relieving pain. The amino acid sequence of the oligopeptide disclosed by the invention comprises a derivative sequence formed by additionally arranging 1-3 amino acids at the N end and / or C end of the amino acid sequence shown as Tyr-Pro-Arg, especially Tyr-Pro-Arg-Arg (YPRR, YR4), and a gastrointestinal tract digestion product Tyr-Pro-Arg (YPR, YR3) of the oligopeptide. The oligopeptide not only can relieve acute pain induced by a mouse thermal drifting model, but also can relieve chronic pain induced by a rat sciatic nerve chronic compression injury model and a tibia bone cancer pain model, and meanwhile, the oligopeptide obviously inhibits CCI-induced spinal astrocyte, microglia and neuron activation; therefore, the compound has important development and application values in the aspects of central sensitization inhibition and analgesia.

The present invention discloses an adversarial sample detection method and device based on a neuron activation map, which obtains an image data set to construct and train a deep model; generates noise samples and constructs a binary data set X B ; reads the checkpoint file saved by the deep model f generated in step (1), and generates the corresponding binary data set X B and the corresponding mapping weight set W B ; the mapping of the mapping weight set W B to the graph set G B ; calculates the corresponding weighted degree set D B of the graph set G B ; trains an adversarial sample detector; during application, inputs the to-be-detected image sample into the trained adversarial sample detector, and after inspection, outputs whether the to-be-detected image sample belongs to a clean sample or an adversarial sample, thus realizing the detection of the to-be-detected image sample.

The invention provides a large model safety neuron screening method and device, and the method comprises the steps: carrying out the safety alignment of a basic large model, and obtaining a safety alignment model; calculating a neuron activation difference between the security alignment model and the basic large model; and based on the neuron activation difference, determining a safe neuron when safe alignment is performed on the basic large model. The method starts from the internal properties of the model, is not limited by task forms, is easy to expand, is suitable for finding safe neurons in the large model, and provides a scheme for further researching the safety mechanism of the large model.

Systems and methods for clinical neuronavigation in accordance with embodiments of the invention are illustrated. One embodiment includes a method for generating a brain stimulation target, including obtaining functional magnetic resonance imaging (fMRI) image data of a patient's brain, where brain imaging data describes neuronal activations within the patient's brain, determining a brain stimulation target by mapping at least one region of interest to the patient's brain, locating functional subregions within the at least one region of interest based on the fMRI image data, determining functional relationships between at least two brain regions of interest, generating parameters for each functional subregion, generating a target quality score for each functional subregion based on the parameters and selecting a brain stimulation target based on its target quality score and the patient's neurological condition.

The present application relates to the technical field of automatic driving model test verification, and is a hardware-in-the-loop automatic driving model vulnerability mining method and system, specifically comprising: in the hardware-in-the-loop closed loop operation, through the non-invasive probe deployed in the tested chip, the neuron activation features of the model hidden layer are extracted, and the cognitive uncertainty index is calculated; when the uncertainty exceeds the threshold value, the physical parameters such as illumination, weather, shielding and target geometric shape are dynamically adjusted, and the targeted adversarial negative sample scene is generated; in the scene evolution process, the algorithm output, the model internal state and the chip hardware state are monitored synchronously, the failure reason is determined through multidimensional comparison, and the vulnerability is classified into algorithm logic weakness, quantization precision weakness or hardware performance weakness. The present application carries out active exploration based on cognitive driving, realizes efficient and accurate mining of model vulnerability and finding of software and hardware coupling failure, and is especially suitable for algorithm robustness evaluation of domestic heterogeneous computing platforms.