91 results about "Work memory" patented technology

The invention discloses a clinical intelligent decision-making method based on proxy workflow and a storage medium. Comprising the following steps: acquiring and preprocessing clinical information of a patient, and constructing a candidate disease set; and constructing a proxy directed workflow. In the retrieval stage, the diagnosis criteria corresponding to the candidate diseases are retrieved and aggregated from the diagnosis criteria library rechecked by the experts to form working memory. The preliminary diagnosis stage model node generates a preliminary candidate diagnosis set in combination with work memory and patient medical history and physical examination. And the final diagnosis stage generates a final diagnosis result based on the preliminary candidate diagnosis and the complete clinical information. And when the global confidence is lower than a threshold value, the model node pointedly checks an information source and updates reasoning and confidence. A successful reasoning track forms a demonstration set after manual auditing, the demonstration set is used for supervising a fine tuning model to obtain an initial strategy, multiple structured outputs are generated through grouping sampling, relative strategy updating is carried out in combination with reward signals and reference strategy regularization constraints, and optimization and stable improvement of the model diagnosis capability are achieved.

A multi-agent and human project management and collaboration tool can provide the ability to create and manage projects and software agents in a cohesive framework. The collaboration tool can provide a multi-agent, multi-player channel where agents and humans work together to achieve an outcome. The collaboration tool can initiate a new project, where the objective is defined, and a team is dynamically built to achieve this objective. The team roles are parsed, and an agent is created for each role. Each agent is equipped with a personalized skillset and a working memory, allowing them to effectively collaborate with humans and each other to progress autonomously through the project stages. The agents can complete tasks based on the project plan and the results are stored in the agent's working memory. The project status is updated, and a report can be generated after each stage, providing a record of the project's progress.

The invention discloses a robot training system and method based on plot memory in the technical field of artificial intelligence and robot learning, and solves the problems that an existing robot training method lacks an effective memory scheduling system, an empirical value quantification mechanism is not intelligent, and the multi-modal information fusion capability is insufficient. The system comprises a sensing module, a multi-modal unified memory encoder, a progressive memory scheduling system, a multi-dimensional value quantitative evaluation mechanism, an intelligent experience hierarchical scheduler, a multi-modal unified code retriever, a strategy generation module, an execution module and an anomaly detection module. The multi-modal unified memory encoder adopts a hierarchical dimensionality reduction Transform architecture, and fuses an RGB image, a depth image, force sensor data and joint angle information into a 576-dimensional unified feature vector; the progressive memory scheduling system comprises a working memory structure, a short-term memory structure and a long-term memory structure. The multi-dimensional value quantitative evaluation mechanism carries out quantitative scoring on experience based on reward evaluation, novelty evaluation and uncertainty evaluation.

The invention provides a body navigation method, device and equipment based on brain inspiration structured spatial memory, and relates to the technical field of navigation.The method comprises the steps that a work memory module is used for determining a structured spatial memory module for searching a navigation instruction according to the complexity of the received navigation instruction, and the structured spatial memory module is used for searching the navigation instruction; the structured spatial memory module comprises at least one of a landmark memory module and a cognitive map module; responding to the navigation instruction, and performing hierarchical retrieval by using a structured spatial memory module of the navigation instruction to obtain a candidate coordinate set of the target; the candidate coordinate set comprises a plurality of candidate coordinates; determining an exploration sequence based on the confidence score of each candidate coordinate and the distance between each candidate coordinate and the coordinate of the current position of the body agent; the exploration sequence represents the exploration sequence of the candidate coordinates; and controlling the body agent to execute navigation operation based on the exploration sequence. According to the method, the robust space cognition and self-adaptive navigation capability of the intelligent agent with the body in a complex environment is realized.

A system for measuring the surface roughness of surfaces machined by electrical discharge machining (EDM) using a convolutional neural network, comprising: an image recording device with a camera module for recording images of a machined surface with depressions and burrs produced by electrical discharge machining; an image preprocessing module that is connected to the image capture device and is configured to improve the texture contours of the captured images and optimize the image; a Convolutional Neural Network module configured to receive images captured by the image preprocessing module, wherein the Convolutional Neural Network module comprises the following: a feature extraction submodule with multiple convolution and pooling layers, wherein this submodule is configured to analyze the distribution of light intensity variations in the acquired images, where darker areas correspond to valleys and lighter areas to edges of the processed surface, and extracts features from this; and a classification submodule with fully connected layers and an output layer, wherein the classification submodule is configured to predict an arithmetic mean roughness value (Ra) based on the extracted features; a computer device comprising a processor, a working memory and a storage unit, wherein the computer device is configured to execute the convolutional neural network module using the processor, the execution parameters being loaded from memory into the working memory; and an output module integrated into the computer device, wherein the processor is configured to create a database of samples with recorded surface roughness values ​​within the storage unit, the database containing the preprocessed image with timestamp and image metadata as well as the results obtained from the prediction, and wherein the processor is further configured to deliver the predicted arithmetic mean roughness value (Ra) from the output layer to a user interface.

In various examples, temporal memory fusion in scene perception for autonomous and / or semi-autonomous systems and applications is described herein. Systems and methods are disclosed that use a temporal fusion model (e.g., the model) to generate representations—such as BEV images-associated with an environment. For instance, a working memory buffer may maintain working memory features, such as in a fixed-lag manner. A temporal overlap heatmap—such as a temporal overlap image—may then be temporally propagated to indicate a number of times individual grid cells fall within fields of view associated with temporal features. As such, to generate a representation using the model's temporal reasoning capabilities, the model may fuse the heatmap with the working memory features and / or the current features using one or more neural networks. The model may then process the fused output using one or more decoders to generate the representation.

The invention provides a maze robot brain-like cognitive system based on an emotion and memory mechanism, and the system comprises a perception and emotion generation module which obtains the environment state of a robot, and calculates the reflective fear value and cognitive fear value of the robot; the memory module is used for recording short-time memory, working memory and long-time memory of the robot in combination with an anti-thought memory mechanism; the behavior decision-making module is used for comparing the evaluation function based on the cognitive fear value with the energy margin to obtain a behavior decision-making result; the learning module is used for exploring and learning a target node and an energy node to obtain an updated target guide table and an updated energy guide table; and the action selection module is used for updating the target guide table and the energy guide table according to the behavior decision result, and outputting the next action. According to the invention, the survival rate of the robot in the environment learning process can be improved, and long-time learning is realized; and the robot can be prevented from repeatedly exploring in an invalid area, the learning efficiency is improved, the energy consumption is reduced, and self-survival is ensured.

The invention discloses a network equipment configuration file verification method and system based on a Drools rule engine, and the method comprises the steps: receiving a to-be-verified network equipment configuration file, carrying out the line-by-line analysis of the configuration file, and constructing a configuration line object; a configuration type mapping table is maintained, and a multi-stage matching strategy is adopted for configuration type identification; a Drools rule engine is loaded, a KieServices instance is created, and grouping management of verification rules is supported; inserting the analyzed configuration line object as a fact into a working memory of a Drools rule engine, performing verification rule matching and execution, and collecting a detailed verification result at the same time; and after verification is completed, statistical analysis and formatting processing are performed on a verification result, and a report is generated. According to the method and system, verification rules and program logic are separated, and dynamic management and execution of the rules are achieved through a Drools rule engine.

The invention relates to the technical field of artificial intelligence question and answer interaction, in particular to an agent question and answer interaction method based on a large model and generative artificial intelligence, and the method comprises the steps: screening out an information unit packet which needs to be stored in a short-term memory based on the time interval between each information unit packet and question content and the similarity of embedded vectors; obtaining the time delay of the intelligent agent at the current moment, and dynamically adjusting the capacity threshold value of the short-term memory in combination with the distribution of importance scores of all information unit packets in the short-term memory; and answering the question content based on the information unit packets stored in the working memory, the short-term memory and the long-term memory of the intelligent agent at the current moment. According to the method, the problems of long dialogue key information loss, cross-dialogue memory interruption and excessive context information in a traditional intelligent question and answer interaction method are solved, and the interaction continuity, the response speed and the question and answer interaction stability in long dialogue and cross-dialogue scenes are improved.

The invention discloses a large language model vulnerability testing method based on a working memory tree. The method comprises the following steps of 1, collecting malicious semantic texts; 2, constructing an automatic confrontation prompt; 3, testing vulnerabilities of the large language model; 4, evaluating a test result; and 5, carrying out iterative optimization on confrontation prompt. According to the method, antagonism prompts with concealment and attack diversity can be accurately constructed, and the limitation that a traditional large language model vulnerability testing method is limited in coverage range and insufficient in multi-prompt exploratory performance is improved.

A computer-implemented method for performing memory planning for a code generation for determining a code for computing a neural network includes (i) providing successive calculation steps of the neural network, wherein for each calculation step the size of an input data block and an output data block is determined, (ii) determining a condition for the memory planning for each specific computing step, in which an input data block in a memory area is at least partially contained in the output data block, wherein the condition indicates that the memory area associated with the output data block of the calculation step preceding the specific calculation step is contained in the memory area of the output data block of the specific calculation step, and (iii) performing memory planning in which the memory area of the respective input data block and output data block in the working memory is determined for each calculation step, taking into account the determined conditions.

The application discloses a product service network simulation platform based on multiple agents, and belongs to the field of service computing and simulation platform. The simulation platform comprises an agent template library, a digital twin interface, a topology manager and a simulation controller. The agent template library provides agents for representing roles in the product service network, and configures a predefined element attribute template for each agent. Each agent further comprises an agent memory set, which is divided into a short-term working memory, a long-term fact memory, a situational memory and a reflective memory. The digital twin interface acquires and records external simulation environment data in real time or periodically, ensures that the historical data is auditable, and inputs the data into the agents. The topology manager is used for mapping the interaction relationship of the roles in the product service network into the interaction relationship between the agents. The simulation controller is used for simulating each agent. The application breaks through the bottleneck of computational efficiency and scalability, and reduces the technical threshold of complex simulation construction and analysis.

Mechanisms are disclosed for detecting one or more defects in a system In-system and externally observable events are received. Active state information is derived from the in- system and externally observable events. If a memory update is required based on facts from the semantic memory, an episodic memory and a working memory are updated. Actions required from a procedural memory are extracted based on the updated episodic memory and the updated working memory. A prompt is executed on a generative language model to produce a model reasoning output. Reproduction steps are derived from the episodic memory, based on a determination that a defect is present from the model reasoning output. A defect signal is generated and provided to a testing user.

A system and method are disclosed for securely transferring or storing protected working memory outside an owner thread while maintaining full encryption throughout its lifecycle. The protected working memory has an encrypted data component and a keystream component containing dynamically updated encryption information used for per-cycle decryption and re-encryption. A hardened cryptographic subsystem, such as a Trusted Execution Environment (TEE), Secure Element (SE), TPM, or HSM, performs authenticated encryption using device-bound key material to wrap either (i) the keystream component alone or (ii) the entire protected working memory blob. The wrapped object is suitable for storage or transfer across threads or in external memory, without exposing plaintext or keystream material. When restoring, an authorized owner thread supplies a key handle to the hardened subsystem to securely decrypt and reconstruct the protected working memory only within a guarded heap. At no point are plaintext contents or keystreams available to the operating system or user-accessible memory.

The invention relates to a computer-implemented method for performing memory planning for code generation to determine a code for the computation of a neural network in a hardware environment (2), comprising the following steps: - Providing (S1) successive computation steps of the neural network, wherein several of the computation steps provide for the computation of a weight prefetching layer for which weight prefetching is applicable; - Assigning (S3) a possible weight prefetching to at least some of the multiple weight prefetching layers to obtain several different combinations of weight prefetching layers, wherein the weight prefetching involves preloading network parameters into a working memory of the hardware environment for the respective weight prefetching layer; - Performing (S4) memory planning for the code to be created for the several different combinations, wherein the respective memory planning for the successive computation steps of the neural network is carried out taking into account the respective combination of the weight prefetching layers for which weight prefetching is provided, wherein for the respective memory planning for each computation step the memory area of ​​the respective input data block, output data block and at least one network parameter block in main memory is determined and a maximum required memory space of main memory is determined; - Select (S5) one of the memory schedules depending on the maximum required memory space.

This disclosure relates to methods and corresponding apparatus for managing convolutional computations. In one embodiment, a method for managing convolutional computations performed by a computing unit adapted to compute output data on an output channel based on convolutional kernels applied to input data blocks on at least one input channel, wherein computation for each input data block corresponds to output data on an output channel, and wherein computation for each convolutional kernel corresponds to output data on each output channel, the method comprising: identifying the size of a memory location available in a temporary working memory of the computing unit; preloading a maximum number of convolutional kernels that can be stored within the memory size in the temporary working memory; and controlling the computing unit to compute a set of output data that can be computed based on the preloaded convolutional kernels.

The invention discloses a self-adaptive illumination system and method for promoting cognitive efficiency based on task cognitive types, and relates to the technical field of intelligent illumination, and the method comprises the steps that a multi-mode task recognition module obtains user behavior data in an intelligent self-adaptive mode, extracts behavior characteristics of the user behavior data, and judges the task cognitive types of the user based on the behavior characteristics; the central processing and control module obtains an optimal lighting parameter through a reinforcement learning algorithm based on a task cognition type and user basic information in an intelligent self-adaptive mode, and sends a control instruction to the dimmable lighting module based on the optimal lighting parameter; the dimmable lighting module receives the control instruction and adjusts lighting parameters of the lamp according to the control instruction; the on-line efficiency verification and strategy optimization module quantifies the working memory efficiency of a user in the current light environment based on a micro-task in an intelligent adaptive mode, feeds back a working memory efficiency evaluation result to the central processing and control module, carries out algorithm optimization, and dynamically provides an optimal light environment capable of directly improving the working memory efficiency.

The optical-electro-mechanical system for enhancing the working memory comprises a transcranial supporting module, a phototherapy module, a blood oxygen and heart rate detection module and a main control module, and the phototherapy module is used for emitting treatment light waves to the head of a user; the blood oxygen and heart rate detection module is used for detecting blood oxygen and heart rate of a user and comprises the steps that PPG signals of the user are collected, a signal quality score is calculated, when the signal quality score is smaller than a score threshold value, if the absolute value of the light intensity amplitude difference value of a current sampling point and a previous sampling point of the PPG signals exceeds a first threshold value, the current sampling point is judged to be an abnormal point, and the current sampling point is judged to be an abnormal point. Repairing the current sampling point, otherwise, performing adaptive Kalman filtering on the PPG signal; calculating a blood oxygen value, a heart rate value and a standard deviation of an RR interval of the user according to the filtered PPG signal; the main control module is used for adjusting the light intensity, the duty ratio and the treatment time of the treatment light waves output by the phototherapy module according to the blood oxygen value, the heart rate value and the RR interval of the user.

The invention relates to a computer-implemented method for executing a storage plan for determining code for computing a neural network for use in a hardware environment (2), comprising the following steps: providing (S1) successive computing steps of the neural network, wherein the size of at least one input data block (EB1, EB2, EB3, EB4) and at least one output data block (AB1, AB2, AB3, AB4) is determined for each calculation step; -determining (S2), for each calculation step, a maximum overlap region between the input data block (EB1, EB2, EB3, EB4) and the output data block (AB1, AB2, AB3, AB4); and performing (S3) a storage planning in accordance with the maximum overlap region for each calculation step, wherein the storage regions of the respective input data blocks (EB1, EB2, EB3, EB4) and output data blocks (AB1, AB2, AB3, AB4) are specified in the working memory.

The invention discloses a data processing method and device. The method comprises the following steps: acquiring a tool calling result generated by an interaction system according to client input; inserting a first content identifier in the tool calling result to obtain working memory data of the interactive system, and storing the working memory data to a context content cache; and replacing context content data in the working memory data with a content placeholder, wherein the content placeholder is used for indicating the interactive system to recover the context content data from the context content cache according to the first content identifier. According to the scheme, the context content data is marked and cached, and the context information is quickly recovered by using the first content identifier in the subsequent calling, so that efficient memory management and information reuse of the interaction system in multiple rounds of interaction are realized, the response speed is effectively improved, the repeated calculation amount is reduced, and the context consistency and the context integrity are ensured.

The invention provides a composition for improving learning memory, spatial memory and / or cognitive functions and application thereof. The composition disclosed by the invention comprises DHA (docosahexaenoic acid) and galactose, and the mass ratio of the DHA to the galactose is 1: (8-260). The invention also provides a food containing the composition, and an application of the composition or the food in preparation of a product helpful for improving spatial memory, working memory and / or cognitive function.

The invention relates to a chiplet architecture (2) for a technical system (1), comprising the following chiplets (3): - a system management controller (4) of the technical system (1), - an artificial intelligence hardware accelerator (5) for applications, - a central processing unit (6) for applications of the technical system (1), - an internal graphical processing unit (7), and - a working memory (8).

The application relates to the cross field of biomedical engineering and artificial intelligence, and discloses a personalized brain power development training method based on electroencephalogram signals. The method comprises the following steps: collecting resting state and task state multi-channel electroencephalogram signals of a subject, constructing a functional connection matrix after pretreatment, identifying individualized weak connection target points through difference operation and cluster analysis; matching a neural feedback training protocol from a preset paradigm library based on the target points, and generating a feedback signal by extracting a target point synchronicity feature in real time during training to guide the subject to actively enhance the weak connection; updating the model after each training and dynamically optimizing subsequent parameters to form a closed-loop regulation. The application improves working memory and attention through individualized targeted training, induces neural plasticity, and realizes efficient and accurate brain power development.

The invention discloses an ICU (Intensive Care Unit) long-time-sequence decision-making auxiliary method based on hierarchical memory and dynamic abstract, which comprises the following steps of: organizing electronic health record data of a patient into a hierarchical memory architecture which comprises instantaneous working memory, abstract scene memory and archiving retrieval memory, and when a diagnosis and treatment query is received, storing the data into a database; a dynamic prompt assembler dynamically assembles a final prompt by extracting and retrieving data from three memory hierarchies, and finally submits the final prompt to a large language model to generate a diagnosis and treatment suggestion. According to the method, the cognitive process of clinical experts can be simulated, structured and hierarchical management can be performed on ICU long-time-sequence data, and latest dynamic and medium-term trends and key historical events are intelligently integrated when clinical query is responded, so that the problems of context length limitation and information submergence confronted when LLM processes long-time-sequence clinical data are fundamentally solved.

Methods and systems for determining a state of working memory of a mammalian brain are disclosed. An example method includes positioning a device on an individual and obtaining measurement data of the individual's brain. The measurement data includes time series data collected from sensors over a period of time including at least one stimulus event. Temporal information of the stimulus event is obtained. A spectrogram of the measurement data is computed for each of the sensors. A classifying technique is used to identify onsets of activation and durations of activation within at least one frequency sub-band of the spectrogram in time. A state of working memory is determined based on the onsets and durations identified. The state of working memory temporally correlated with the stimulus event is output. Methods and systems as disclosed may be useful for quantitatively or qualitatively evaluating cognitive performance of a mammalian brain.

The invention relates to a computer-implemented method for executing a memory plan in a hardware environment for code generation for determining code for neural network calculations, comprising the steps of: providing successive calculation steps of layers of a neural network, determining the size of the input data block, the size of the output data block and the size of one or more model parameter blocks according to the type of the calculation step, wherein the one or more model parameter blocks have model parameters for the corresponding calculation step; determining a rule for memory planning for each particular calculation step that requires the use of the model parameters, wherein the rule specifies: loading the model parameters into contiguous memory regions in a working memory of the hardware environment; a memory plan is executed, in which memory regions of the respective input data block, output data block and model parameter block in the working memory are ascertained for each calculation step taking into account the determined rules.

This invention discloses an embodied intelligent agent and a method for ensuring its operational security. The method includes: constructing a hybrid long-short-term security memory for the embodied intelligent agent, comprising a long-term security memory storing cross-task security experiences and a short-term working memory recording the current task's temporal trajectory; performing forward predictive reasoning based on the long-term security memory to retrieve security experiences relevant to the current context and perform contextual verification on candidate actions output by the embodied intelligent agent; triggering an interception mechanism to prohibit the execution of candidate actions when a contextual risk is detected; and performing backward reflective reasoning based on the short-term working memory to infer temporal security constraints during task execution by analyzing global task instructions and verifying temporal risks by tracing back historical trajectories; triggering a replanning mechanism and generating a corrective action sequence when a temporal risk is detected. Thus, by designing bidirectional security reasoning, the security of the embodied intelligent agent's task execution is effectively improved.

System for diagnosing a battery cell (204) of a vehicle (100), comprising: a processor (302) and a working memory (304) or data storage (308) comprising an algorithm or computer instructions which, when executed by the processor (302), perform an operation comprising: Determine that a current flows through a current collector (206, 214) of the accumulator cell (204), Generating a mechanical excitation of the current collector (206, 214), Determining the amplitude of a voltage across the accumulator cell (204) based on mechanical excitation, and Determining the presence of a crack or separation of a foil of the current collector (206, 214) on the basis of the amplitude of the voltage across the accumulator cell (204).

System (100), comprising: a working memory that stores components executable by a computer; and a processor that is operatively connected to the main memory, which executes computer-executable components, wherein the computer-executable components include: a patch identification component (116) that identifies a workload which belongs to a set of patches which belongs to a hybrid cloud-based data processing platform, wherein the workload has a first workload part for which a first subset of patches is to be performed, and a second workload part; a patch management component (114) that: a second subset of patches is determined, which is to be executed with respect to a part of the first workload part based on hazard assessments of patches of the first subset of patches within a maintenance window, wherein the first subset of patches includes the second subset of patches, a third subset of patches is determined to be executed with respect to the second workload part, wherein the set of patches comprises the first subset of patches, the second subset of patches, and the third subset of patches; and a patch execution component (118) that executes the second subset of patches with respect to the part of the first workload part within the maintenance window while the first workload part of the cloud-based hybrid computing platform is offline, and executes the third subset of patches with respect to the second workload part while the second workload part of the cloud-based hybrid computing platform is online.

The invention provides a visual reasoning method based on reflection mechanism working memory. The visual reasoning method comprises a visual feature analysis step, a working memory coding step, a reasoning step and a result output step. According to the invention, work memory optimization based on a reflection mechanism is realized through work memory noise addition and long-term memory guide denoising, and effective interaction between work memory and long-term memory is realized, so that the work memory can be self-adjusted and corrected by using stable knowledge in the long-term memory; the long-term stability and the information retention capability of the system are enhanced, so that the system maintains stable memory representation in the visual reasoning task and is not easily interfered, and the self-adjusting capability in the visual reasoning task process is enhanced. A flow matching network is introduced to realize a denoising process, and an optimization process of a working memory is constructed into a continuous-time dynamic system, so that the working memory can be gradually recovered to a noiseless state along an optimal track, and the robustness and adaptability of memory representation are enhanced.