27 results about "Memory retention" patented technology

The invention discloses a memory optimization system for a large-scale language model, and the system comprises a dynamic semantic memory modeling module which is used for generating a context-aware memory vector through a bidirectional Transform, and supporting the increment to be updated to one hundred million entity records; the self-adaptive memory resource scheduling module dynamically allocates memory pool partitions based on inference load, high-frequency memory is stored in a working area, and quantitative compression is started in an archiving area; the memory attenuation and enhancement module adopts a two-factor attenuation mechanism to execute enhanced storage and regular pruning on the high-value memory; the cross-scene memory migration module is used for realizing cross-domain adaptation through a domain adapter and comparative learning; and the memory conflict resolution module identifies conflicts based on the rule base and adopts a multi-factor voting mechanism to resolve the conflicts. According to the method, the memory retention integrity and the retrieval efficiency are improved, the resource utilization is optimized, the cross-scene adaptation capability is enhanced, the memory conflict is eliminated, and the model reliability is improved.

The embodiment of the invention discloses a communication method, device and equipment between Jailhouse virtual machines and a medium, and the method comprises the steps: adding a VNIC module in a Jailhouse source code, and enabling the VNIC module to analyze VNIC configuration and achieve address space mapping, thereby achieving the management of the VNIC; modifying a device tree file of each virtual machine operating system, and adding a memory retention statement which is used for address mapping of the VNIC; driving compatible information is added in the device tree file of each virtual machine operating system, a virtual area mapped by the VNIC is registered based on the memory reservation declaration, and a VNIC interrupt signal is configured; setting a VNIC MAC address in the device tree file according to the number of the virtual machine, and setting a specific bit of the MAC address as the number of the virtual machine for identifying a node to which the communication data flow belongs; and according to the VNIC interrupt signal of the device tree file, modifying an interrupt signal in a memory registration structure body in the configuration file to be consistent with the VNIC interrupt signal of the device tree file, and correspondingly setting an actual memory region mapped by the virtual region.

The invention relates to a dictionary pen learning assistance method and device, equipment and a medium, and the method comprises the steps: extracting learning entries inquired by a user in a target time sequence, and determining the user features and word attribute features of each learning entry; constructing a multi-dimensional feature vector of each learning entry; substituting the multi-dimensional feature vector into an established forgetting curve model, and calculating a corresponding forgetting state; according to the forgetting state, a corresponding word learning prompt strategy is dynamically generated, and the word learning prompt strategy is used for automatically pushing corresponding review prompt information at the target review time point; and obtaining a response behavior of the review prompt information, determining a corresponding state field according to the response behavior, and updating a corresponding word learning prompt strategy according to the state field. By introducing the forgetting curve model to dynamically generate the learning prompt strategy, intelligent review recommendation for the personal memory rhythm of the user can be realized, so that the learning efficiency and the memory keeping effect are improved.

According to the real-time multi-level tile map synthesis method provided by the invention, five-level processing pipelines are not simply parallel, and independent segmentation factor level design, lock-free cross-level data channels and three-level synchronous barrier architecture are carried out; a dynamic memory release mechanism driven by a global tile counting dictionary and a boundary tile divide-conquer synthesis algorithm are used. Through global tile counting dictionary monitoring, the memory is released once the tiles are completed, and accumulation and occupation of the memory are avoided; a boundary adaptive threshold value avoids memory retention caused by the fact that boundary tiles wait to be filled up; five layers of independent treatment pipelines are designed, and the actually measured treatment speed is increased by 4-5 times compared with that of a single-stage serial system; synchronous barriers are arranged in the three stages of initialization, real-time processing and termination, and data consistency is ensured; 1: 1 to 1: 64 times of dynamic scaling is supported through five-level segmentation factors, original features of the image are kept while scaling is carried out by using a nearest neighbor interpolation algorithm, and coordinate range inspection is carried out.

The utility model discloses a mind mapping sounding device for a cardiac pacemaker implanted patient, relates to the technical field of medical auxiliary equipment, and aims to solve the problems that in the prior art, traditional patient education depends on written materials or oral explanation, information is fragmented, the memory retention rate of the patient is low, and postoperative follow-up visit content is complex and is difficult for the patient to remember. Comprising a flexible map guide board and a control box, the upper end of the flexible map guide board is provided with a middle node, a first node and a second node, one side of the first node and one side of the second node are provided with a touch sensing area, the control box is provided with a prompt lamp and a voice interaction module, and the voice interaction module comprises a voice output module and a voice input module. A master control module, a storage module, a data interaction module and a follow-up time management module are further installed in the control box. Operation knowledge, postoperative care, follow-up visit nodes and the like are integrated through mapping, the cognitive level and follow-up visit compliance of a patient are remarkably improved, and the occurrence rate of complications is reduced.

The invention discloses a learning planning method. The method comprises the following steps: acquiring the correct rate of each evaluation dimension of a knowledge point corresponding to an nth review test event; according to the importance of each evaluation dimension, fusing each correct rate to obtain a comprehensive mastering degree P corresponding to the nth review test event; calling a time interval tn; performing memory ability analysis processing based on the comprehensive mastering degree P and the time interval tn to obtain memory intensity S corresponding to the nth review test event; acquiring a preset memory retention rate R; and carrying out time information conversion processing based on a preset memory retention rate R and the memory intensity S corresponding to the nth review test event to obtain a suggested interval duration of the (n + 1) th review test event. According to the method, the problem that the review effect is gradually reduced due to the fact that the current learning plan deviates from the memory condition of the user can be solved, and the requirements of partial emphasized dimensions are also considered while efficient review is carried out on each dimension according to the memory condition of the user.

The present application provides a kind of nutritional composition in the application of the product for improving learning, memory and cognitive ability.The nutritional composition includes animal bifidobacterium lactis subsp.CP-9 and 3'-sialyllactose.Animal bifidobacterium lactis subsp.CP-9 and 3'-sialyllactose can synergistically enhance the memory retention ability of the body, reduce the degree of learning and memory errors, and up-regulate the expression of oligodendrocyte transcription factor 2 coding gene, myelin basic protein coding gene and brain-derived neurotrophic factor coding gene, realize the double regulation of myelin formation and neurotrophic in hippocampus region of brain, synergistically improve the neural plasticity of central nervous system through gut-brain axis pathway, and then double improve the learning, memory and cognitive function of the body from the behavior and molecular level.

The invention provides a method for training the visual clue-dependent positioning reward capability of animals by using an animal behavior training and testing system, and the system realizes that one set of system covers multiple types of experiments through the opening and closing logic of an access door of an adjusting arm, the reward region reward putting rule and the free conversion of visual clues. A visual clue-reward association learning scene is constructed, and animals are guided to establish memory association between clues and reward positions. A three-stage visual clue navigation and memory ability testing method is established, the memory retention ability associated with clue rewards is evaluated in the association stage, the spatial memory ability is independently evaluated in the blank stage, and the regulation and control mechanism of visual clues on animal spatial memory is analyzed in the non-association stage.

The application discloses a memory optimization system for a large language model, comprising: a dynamic semantic memory modeling module, which generates context-aware memory vectors through a bidirectional Transformer and supports incremental updating to 100 million entity records; an adaptive memory resource scheduling module, which dynamically allocates memory pool partitions based on inference load, stores high-frequency memories in the working area, and enables quantization compression in the archive area; a memory decay and enhancement module, which adopts a double-factor decay mechanism, performs enhanced storage on high-value memories, and prunes them regularly; a cross-scenario memory migration module, which realizes cross-domain adaptation through a domain adapter and contrastive learning; and a memory conflict resolution module, which identifies conflicts based on a rule base and resolves contradictions using a multi-factor voting mechanism. The application improves memory retention integrity and retrieval efficiency, optimizes resource utilization, enhances cross-scenario adaptation capability, resolves memory conflicts, and improves model reliability.

The application provides a large model knowledge dynamic iteration updating method and system, relates to the technical field of large models, collects multiple types of dynamic knowledge sources in the field of children's tumors and is attached with relevant information, and constructs a knowledge evolution life cycle network based on this; the knowledge content nodes are subjected to multi-source credibility cross verification and dynamic conflict reconciliation processing to obtain a target knowledge set; a parameter fine tuning and memory reservation collaborative mechanism is called to convert into a knowledge updating parameter package; the knowledge updating parameter package is input into a large model to perform collaborative updating operation and generate a knowledge updating whole-process traceability document. The application can comprehensively and accurately process multiple types of knowledge sources, effectively improve the knowledge accuracy and reliability, and realize model collaborative updating.

The application belongs to the field of intelligent teaching, and particularly relates to an English oral pronunciation correction system based on artificial intelligence, which comprises a multi-modal perception module, a deep pronunciation feature extraction module, a cognitive compensation fusion module and an intelligent correction and training planning module; the abstract pronunciation is converted into a tactile experience through a multi-modal learning process, the tactile information and the auditory information are synchronously input, the memory retention rate of the visually impaired students is improved, and the students are guided to deeply remember the acoustic characteristics of the pronunciation; the application simulates the auditory-tactile joint representation ability of the visually impaired brain through the learnable attention weight and the cognitive compensation prior mask, realizes the real sense of "touch instead of vision", simultaneously identifies the error reasons through multi-task learning, provides scientific basis for targeted training, designs a special tactile generation network, maps the acoustic characteristics and the pronunciation error types into tactile patterns, provides instant tactile correction feedback for the students, and forms a closed-loop multi-modal learning experience.

The invention provides an English vocabulary cognition training method and system based on an interval repetition memory method, and relates to the technical field of language cognition training, abnormal value elimination and normalization preprocessing are performed on collected data, a memory intensity coefficient is calculated, a memory intensity level is judged, and a final review interval is generated in combination with a vocabulary basic interval and a special adjustment rule; each layer is matched with a differentiated multi-modal training task, and the vocabulary adding proportion, cognitive layer dynamic switching and fatigue relieving links in the training process are controlled. According to the method, multi-dimensional data such as learning duration, memory performance and vocabulary difficulty and dynamic review intervals are fused, compared with fixed period review, the one-week memory retention rate of vocabularies is increased to 85%, the two-week memory retention rate reaches 72%, the long-term memory effect is remarkably enhanced, and the problem that multi-dimensional data such as intervals, non-fused reaction time and cognitive states are adjusted only based on correct rate is solved. And personalized and efficient vocabulary cognition training requirements cannot be met.

The application relates to a device monitoring method, device and computer equipment for a substation debugging process. The method comprises the following steps: establishing a source occupation account page for a subscription object, determining a test source takeover or real source recovery event based on real-time message flow, and updating a current occupation source state; when a test stage end signal is received, sending an occupation release probe message to the subscription object occupied by the test source; determining an occupation release result according to whether a real source recovery message is listened to and whether the current occupation source state has been overwritten as real source occupation; classifying and mapping subsequent business messages to determine semantic attribution, generating an explanation result and overwriting the account page; based on the occupation release result and the latest message explanation classification, when a recovery reconstruction mark needs to be added, a real source recovery reconstruction section is constructed and inserted into the original next round of test process. The method can identify and eliminate semantic dislocation caused by source memory retention of the subscription object in the smart substation debugging process, and improve the debugging reliability.

The invention relates to the technical field of education, and discloses a multi-sensory linkage educational digital oil painting system which comprises a knowledge visualization sketch layer, an intelligent auxiliary terminal layer and a multi-sensory linkage assembly layer. The knowledge visualization sketch layer is used for presenting a digital oil painting sketch fused with educational information, and a color theory mark, a stroke skill prompt and a famous painting background knowledge pop-up window are embedded in the sketch; and the intelligent auxiliary terminal layer comprises a touch operation module, a data processing module and an educational resource library. In the invention, the knowledge visualization sketch layer is embedded with education information such as a color theory and stroke skills through dynamic labeling, audio explanation, tactile feedback and scene smell of the multi-sensory linkage component layer are combined, a hippocampus memory area is activated by relying on a right brain image memory advantage and an alpha wave regulation and control technology, and multi-sensory synchronous input is combined, so that a hippocampus memory area is activated. A knowledge conversion path conforming to the Ebbinghaus multi-channel memory model is constructed, and the memory retention rate is remarkably improved.

A distillation enhancement method for class incremental learning comprises the following steps: (1) constructing a class incremental learning baseline model based on sample playback and knowledge distillation, and performing joint training by using current task data and historical memory library samples in each incremental stage; (2) a semantic confusion matching SCM mechanism is introduced, prediction deviation of a teacher model is quantified by constructing a category confusion matrix, soft labels are corrected, and interference of new categories with similar semantics is inhibited; (3) introducing a long-term memory attention distillation (LTMAD) mechanism, and introducing a distillation weight of time distribution index growth according to the category to strengthen the protection of the early category; (4) combining the SCM loss and the LTMAD loss with the original classification loss and the distillation loss to form a multi-objective optimization function; and (5) carrying out model training and evaluation by using a CIFAR100 type incremental learning data set and an Image Net-Subset type incremental learning data set. According to the invention, the memory retention capability of the class incremental learning model is improved.

The invention provides an intelligent content pushing and interaction method and device, and relates to the field of computers. The method comprises the following steps: constructing a target user portrait based on multi-dimensional user data; based on the target user portrait and the historical learning data, predicting a forgetting time point of each learning unit, and obtaining predicted memory data; wherein in the prediction process, an Ebbinghaus attenuation correction factor is introduced to dynamically adjust predicted data; the Ebbinghaus attenuation correction factor is generated based on the deviation between the actual memory retention rate of the consolidation test of the target user and the theoretical memory forgetting curve; based on the predicted memory data, a grey wolf algorithm is utilized to dynamically optimize multiple combined learning strategies, and an optimal combined learning strategy is determined; and based on the optimal combined learning strategy, establishing a pushing rule, and based on the pushing rule, pushing the corresponding learning content to a user terminal of the target user. According to the invention, intelligent content pushing can be realized, and the real-time performance and accuracy of pushing and the overall performance of the system are improved.

The invention provides an unsupervised cross-modal retrieval method based on memory preserving Hash, and relates to the technical field of unsupervised cross-modal retrieval, comprising the following steps: feature extraction and Hash learning; feature fusion and similar matrix construction; and training and optimizing the model. According to the method, the historical memory similar matrix construction module is used for providing stable training supervision, so that the discrimination capability and retrieval robustness of the Hash codes are effectively improved. According to the method, semantic tags are not needed, the method is suitable for unmarked multimedia data retrieval in a real scene, the model structure is light, the retrieval efficiency is high, and the method has high practical application capacity and expansibility.

The invention discloses an intelligent association review method, a storage medium and equipment, and the method comprises the following steps: S1, constructing a knowledge graph network which comprises a plurality of knowledge point nodes and edges representing the relationship between knowledge points; s2, clustering a plurality of knowledge points to form a knowledge cluster according to a relationship among the knowledge points in the knowledge graph network; s3, calculating a comprehensive forgetting degree of each knowledge cluster, wherein the comprehensive forgetting degree is determined based on a memory retention value and an association weight of each knowledge point in the knowledge cluster; s4, generating an optimal review path according to the comprehensive forgetting degree of each knowledge cluster, wherein the review path comprises a review sequence of the knowledge clusters and a combination mode of associated knowledge points; and S5, dynamically adjusting the review interval of the knowledge cluster according to the actual review effect of the user. According to the method, more scientific memory consolidation can be realized by constructing the knowledge point network, calculating the knowledge cluster forgetting degree and dynamically optimizing the review path.

Certain aspects of the present disclosure are directed towards techniques and apparatus for wireless communication. An example method generally includes: storing one or more communication settings for signal communication in memory; performing a first signal communication using the one or more communication settings; entering an idle mode of operation after performing the signal communication; retaining the one or more communication settings in the memory during the idle mode; and performing a second signal communication after the idle mode using the one or more communication settings retained in the memory.

The envisioned Artificial Intelligence Long-term Memory System presents a significant advancement in artificial intelligence (AI) by addressing the challenge of transient memory in AI language models. This innovation introduces a hardware-centric approach to augment the memory faculties of AI language models, enabling them to store, access, refine, and incorporate specific data points for deeper user engagement. The enhancements focus on long-term memory, facilitating AI models to remember and build upon past interactions, thus offering a more natural and intuitive interaction between computers and users. The potential applications span from personal computing to complex medical diagnostics, marking a pivotal step towards AI models functioning with contextual awareness and memory retention akin to human interaction.

The application relates to a long text memory keeping method, system, device and storage medium based on a large language model. The method comprises the following steps: grading the text content in long text information through a word segmenter in the large language model to obtain various levels of various text contents; respectively calculating the context information weight of the text content at each level based on an attention mechanism to obtain the weight corresponding to each level; storing each text content in a short-term memory space or a long-term memory space according to the weight corresponding to each level; and using a KV Cache to compress the short-term memory space or the long-term memory space, obtaining the text content reserved after compression and outputting the same. The text content in the long text information is graded, and then the weight of the corresponding level is calculated, so that the logical consistency of the context content in multiple dimensions is ensured; the memory space is compressed using the KV Cache, so that the inference time of the large language model can be accelerated.

The application provides an unsupervised cross-modal retrieval method based on memory reservation hash, relates to the technical field of unsupervised cross-modal retrieval, and comprises the following steps: feature extraction and hash learning; feature fusion and similarity matrix construction; model training and optimization. The application provides stable training supervision by using a historical memory similarity matrix construction module, and effectively improves the discrimination ability and retrieval robustness of the hash code. The method of the application does not need semantic labels, is suitable for unlabeled multimedia data retrieval in a real scene, has a light model structure, high retrieval efficiency, and has strong practical application ability and expansibility.

Memory devices that include a pyroelectric layer between two metal layers. The pyroelectric layer may have a crystal configuration that is temperature-dependent. Different crystal configurations have different degrees of polarization. A higher temperature causes a pyroelectric material to have an orthorhombic crystal structure, which has a lesser degree of polarization, leading to lower power consumption and lower leakage. A lower temperature causes the pyroelectric material to have a tetragonal crystal structure, which has a higher degree of polarization, leading to higher power consumption, along with faster switching speed and better memory retention.

The present disclosure provides a communication method. The method includes: receiving an ambient power (AMP) trigger frame for triggering a response from a STA; transmitting, in response to the AMP trigger frame, an AMP response frame comprising a field indicating whether the STA has a capability to support a management mechanism in which the STA has a capability to actively and dynamically switch between an awake state in which the STA performs channel sensing (CS) and a doze state in which the STA performs memory retention (MR) and harvests energy.

The invention discloses a word memorizing device and system based on a forgetting curve and vibration reminding, and aims to improve word memorizing efficiency and optimize a learning process. The device comprises a main control unit, a storage unit, a display unit, an input unit, a vibration reminding unit, a Bluetooth communication module and a power management unit, feedback information of a learner is collected through the input unit, and the main control unit calculates the optimal review time of each word based on a forgetting curve algorithm and updates data in the storage unit; the main control unit further controls the display unit to update display content, learning is reminded through the vibration reminding unit, the device carries out data synchronization with a mobile management terminal through the Bluetooth module, and a lexicon can be managed and review parameters can be adjusted through a mobile terminal application program. Through the forgetting curve algorithm and the vibration reminding function, the learning efficiency and the memory retention rate are effectively improved, and the method is suitable for language learning scenes needing efficient memory review.

This invention relates to the field of learning path planning technology, and provides an adaptive AI learning path planning system based on learning profiles. The system includes a multi-source learning data acquisition module, a data preprocessing module, a three-dimensional cognitive diagnosis module, a multi-dimensional dynamic learning profile construction module, a subject knowledge graph module, an adaptive learning path planning module, and a dynamic path adjustment module. The multi-source learning data acquisition module collects multi-source learning data from students, including basic information data, historical learning data, initial assessment data, and real-time interaction data during the learning process. By integrating dynamic memory states based on the Ebbinghaus forgetting curve, the system can track students' memory retention in real time, ensuring the rationality of the learning sequence, avoiding knowledge gaps, improving learning effectiveness, achieving multi-objective optimization, improving the rationality of path planning, and enhancing the accuracy of cognitive diagnosis, thus providing a more accurate foundation for learning profiles and path planning.

A semiconductor device includes a substrate, a gate oxide layer formed on the substrate, a gate formed on the gate oxide layer, and a spacer formed adjacent the gate and over the substrate. The spacer includes a void filled with air to prevent leakage of charge to and from the gate, thereby reducing data loss and providing better memory retention. The reduction in charge leakage results from reduced parasitic capacitances, fringing capacitances, and overlap capacitances due to the low dielectric constant of air relative to other spacer materials. The spacer can include multiple layers such as oxide and nitride layers. In some embodiments, the semiconductor device is a multiple-time programmable (MTP) memory device.