84 results about "Learning dynamics" patented technology

A technique efficiently routes Internet Protocol (IP) traffic on paths between customer edge devices (CEs) across a provider network (“CE-CE paths”) in a computer network. According to the novel technique, a path computation element (PCE), e.g., a provider edge device (PE), may learn dynamic link attribute information of remote links from the provider network to one or more remote CEs (e.g., “PE-CE links” or “CE-PE links”). A multi-homed requesting CE requests from the PCE a set of CE-CE path metrics (e.g., costs) to one or more remote destination address prefixes, e.g., via each multihomed CE-PE link from the requesting CE. In response to the request, the PCE computes the set of available CE-CE paths and current metrics to the remote destination address prefixes and returns the corresponding CE-CE path metrics to the requesting CE. The requesting CE modifies its IP forwarding entries accordingly in order to perform IP traffic routing corresponding to the CE-CE path metrics (e.g., asymmetrical load balancing) across its multi-homed CE-PE links.

A hierarchical information processing system is disclosed having a plurality of artificial neurons, comprised of binary logic gates, and interconnected through a second plurality of dynamic artificial synapses, intended to simulate or extend the function of a biological nervous system. The system is capable of approximation, autonomous learning and strengthening of formerly learned input patterns. The system learns by simulated Synaptic Time Dependent Plasticity, commonly abbreviated to STDP. Each artificial neuron consisting of a soma circuit and a plurality of synapse circuits, whereby the soma membrane potential, the soma threshold value, the synapse strength and the Post Synaptic Potential at each synapse are expressed as values in binary registers, which are dynamically determined from certain aspects of input pulse timing, previous strength value and output pulse feedback.

The invention provides a learning resource recommendation method, which comprises the following steps: collecting the learning information of a user; preprocessing the learning information to obtain effective learning information; analyzing the effective learning information to extract knowledge points; according to the effective learning information, obtaining the weight of each knowledge point; according to association strength among the knowledge points and the weights of the knowledge points, forming a knowledge point undigraph; and according to the weights of the knowledge points in the knowledge point undigraph, recommending learning resources to a user. The invention further provides a learning resource recommendation system, and the system independently establishes the knowledge point undigraph which conforms to own personality characteristics for different users and records and tracks the learning dynamics of the users in real time so as to improve learning resource recommendation accuracy and improve the learning efficiency and the experience degree of the user.

The invention discloses a traffic flow prediction method based on a global diffusion convolution residual network, and belongs to the technical field of intelligent traffic systems. The method comprises the following steps of: 1, establishing a traffic prediction model based on a global diffusion convolution residual network; 2, learning dynamic correlation and local and global spatial correlation; 3, capturing time correlation and global space-time correlation; and 4, fusing branch results and outputting a final result. According to the traffic flow prediction method, a global diffusion convolution residual network is provided, the model is composed of a plurality of periodic branches with the same structure, and the global attention diffusion convolution network and the global residual network of each branch are used for obtaining the spatial-temporal correlation of each period. Particularly, the global attention diffusion convolution network uses a PPMI matrix based on an attentionmechanism to capture dynamic space-time correlation, and the global residual network uses gating convolution and a global residual unit to capture time correlation and global space-time correlation atthe same time, so that the precision and efficiency of traffic prediction are improved.

A learning apparatus includes a storage unit configured to store a network formed by a plurality of nodes each holding dynamics; a learning unit configured to learn the dynamics of the network in a self-organizing manner on the basis of observed time-series data; a winner-node determiner configured to determine a winner node, the winner node being a node having dynamics that best match the time-series data; and a weight determiner configured to determine learning weights for the dynamics held by the individual nodes according to distances of the individual nodes from the winner node. The learning unit is configured to learn the dynamics of the network in a self-organizing manner by degrees corresponding to the learning weights.

The technology described herein provides a powertrain system with a route-learning feature. Particularly, learned information is used to optimize powertrain operation along any learned route. The learned information comprises, generally, feedback from the vehicle's acceleration and brake sensors and information from an on-board trip computer. At the least, the powertrain is able to optimize its operation to a driver's preference based on the feedback recorded along a particular route that the driver has specified. The route-learning powertrain control described herein is particularly useful with a hybrid powertrain, and can be used to optimize start / stop and regenerative braking control. The system described herein can also be integrated with a navigation system and GPS receiver, to provide more accurate route-learning and / or automated operation.

The invention discloses a personalized recommendation method based on user data of on-line courses. The method comprises following steps of: 1), establishing label mapping relations among courses, labels and titles; 2), calculating an error rate list ET of a user to courses by a course recommendation system according to learning records of the user; 3), generating error lists of all courses according to the list ET and calculating the similarity between two courses according to error lists of courses; and 4), as for each user, calculating prediction weight of error rate of each course of the user according to similarity; calculating recommendation weight of each label according to mapping relations and prediction weight of error rate; calculating recommendation weight of each title according to mapping relations and label recommendation weight and then generating a recommendation list for the user according to recommendation weight of titles. The personalized recommendation method based on user data of on-line courses not only helps to solve a cold start problem but also actively attracts attention of the user in order to improve learning motivation.

The invention discloses a student management system for intelligent teaching. The system comprises a cloud server, a teacher client and a student client, the teacher client and the student client arein signal connection with the cloud server through a network, and the student client comprises a computer host, a display device, a student image acquisition device, an information extraction module,an active analysis module and a memory; and the student image acquisition device comprises at least one group of cameras and is used for monitoring the learning dynamics of students in real time and transmitting a learning state video to the information extraction module. The system has the advantages that teachers and students can thoroughly realize online teaching and learning, and meanwhile, the system capable of monitoring the concentration degree of the students in class in real time is arranged, so that whether the students are in a carefully learning state or not can be judged accordingto the information acquired by the cameras on the student clients, and the students and the teachers can be reminded and prompted in real time.

A self-learning dynamic electrocardiography analysis method employing artificial intelligence. The method comprises: pre-processing data, performing cardiac activity feature detection, interference signal detection and cardiac activity classification on the basis of a deep learning method, performing signal quality evaluation and lead combination, examining cardiac activity, performing analytic computations on an electrocardiogram event and parameters, and then automatically outputting report data. The method achieves an automatic analysis method for a quick and comprehensive dynamic electrocardiography process, and recording of modification information of an automatic analysis result, while also collecting and feeding back modification data to a deep learning model for continuous training, thereby continuously improving and enhancing the accuracy of the automatic analysis method. Also disclosed is a self-learning dynamic electrocardiography analysis device employing artificial intelligence.

The invention discloses a risk stratification method for myocardial ischemia based on deterministic learning and deep learning. The method includes the steps that conventional 12-lead electrocardiogram signals are collected, based on the deterministic learning theory, neural network modeling and identification are conducted on intrinsic electrocardiodynamic characteristics of the shallow electrocardiogram signals, and the intrinsic dynamic characteristics of ECG signals are obtained; the convolutional neural network under the framework of deep learning is used for achieving the risk stratification of myocardial ischemia. The method combines the deterministic learning dynamic modeling method and the deep learning classification method for the first time, the method is applied to early riskstratification of myocardial ischemia based on the conventional 12-lead electrocardiogram signals, no additional detection equipment is needed, and the method is easy and convenient to use and easy tooperate. Through the deterministic learning method, the dynamic characteristics more sensitive to the ischemic state are extracted, the deep neural network can learn data features independently without further data characterization, and the complexity of the system is reduced.

The invention relates to a sensor fusion and improved Q learning algorithm based dynamic barrier avoidance method. The method comprises the steps that S1) the safe distance to a barrier and target coordinate position information and scope during movement of a robot are set; S2) a present pose of the robot is determined, a navigation path is planned, and forwarding is started; S3) in the navigationprocess, environment data detected by a sonar sensor and environment data detected by a laser sensor are preprocessed and characterized and then fused to obtain environment data; S4) whether dynamicbarrier avoidance is needed for the present robot state is determined according to the fused environment data, if YES, a step S5) is carried out, and otherwise, a step S6) is carried out; S5) an improved Q learning dynamic barrier avoidance method is used to obtain the next motion state (a, theta); and S6) whether the robot reaches a target point is determined, if NO, the step S2) is returned to continue navigation, and otherwise, navigation is ended. The method can be used to overcome defects of the single sensor effectively, and improve the barrier avoidance efficiency in the dynamic environment.

The invention relates to a traffic accident prediction method based on a space-time diagram convolutional network, which mainly combines the diagram convolutional network and a long-short-term memory network to respectively obtain space information feature extraction and time sequence information feature extraction of an actual traffic road network, fully considers traffic information data such as time, weather and interest points and constructs a topological structure of a road network by adopting a region division method; firstly, a regional road network is constructed into a graph structure, a graph convolutional network is used for learning a complex road network topological structure to obtain spatial correlation in a traffic state, and then a long-short-term memory network is used for learning dynamic changes of traffic accident data to obtain temporal correlation in the traffic state; and finally, the two networks are combined to construct a prediction model, and the traffic accident is predicted based on the prediction model so as to improve the prediction performance of the traffic accident.

The invention is a queuing message data processing method based on wireless network, comprising the following steps of: an operating hall host computer establishes relation database of logon handheld terminal numbers, reducing codes and queuing numbers; inputting operating hall number and transacted service classified data into handheld terminals; transmitting them through wireless network to the operating hall host computer to log on; the host computer distributes queuing numbers and reducing codes to handheld terminals in sequence and adding them to the database; the computer transmits updated service number, waiting number, time change and other data to each handheld terminal; inputting reducing codes into the host computer, and the computer prints queuing number bill as the basis for service transaction. A transactor can log on the host computer by handheld terminal so as to start queuing and exchanges data with the terminal through the host computer to completely learn dynamic changing queuing conditions of the operating hall.

The present invention discloses a data difference-driven self-learning dynamic batch process optimization method including the following steps: collect production process data off line; eliminate singular batches through PCA operation; construct time interval and index variance matrices to carry out PLS operation to generate initial optimization strategies; collect data of new batches; run a recursive algorithm; and update the optimization strategy. The present invention utilizes a perturbation method to establish initial optimization strategies for an optimized variable setting curve. On this basis, self-learning iterative updating is carried out for mean values and standard differences on the basis of differences in data statistics, so that the continuous improvement of optimized indexes is realized, and thereby a new method is provided for batch process optimization strategies for solving actual industrial problems. The present invention is fully based on operational data of a production process, and does not need priori knowledge about a process mechanism and a mechanism model. The present invention is applicable to the dynamic optimization of operation trajectories of batch reactors, batch rectifying towers, batch drying, batch fermentation, batch crystallization and other processes and systems adopting batch operation.

As part of mobility management messaging in current systems, only static quality of service (QoS) profiles and their statically configured classifiers are reported. These classifiers are known a priori and are typically used for hosted traffic. Once advanced application classifiers are matched and mobility messaging is done, the deep packet inspection (DPI) learned subscriber context is lost. An embodiment of the invention is configured to associate DPI-learned dynamic classifiers with the classifier's policy information as payload over mobility management messaging. The embodiment is useful for identifying protocols and applications, for authentication purposes, solving congestion-based issues between elements in a network, guaranteeing QoS without reserving resources statically, attaching traffic packets to a particular subscriber, and creating mobile signaling to provision end-to-end communication in a network. Thereby enabling the transfer of subscriber parameters and information to be associated with the subscriber throughout subscriber movement through multiple elements of a network.

The invention relates to an induction motor state strong tracking filtering estimation method including parameter tracking. The method is characterized in that a PSO iterative learning dynamic optimization algorithm is adopted to track induction motor parameters on line; and an STF algorithm is adopted to jointly estimate the rotating speed of the induction motor and the rotor flux linkage. According to the invention, the strong tracking filtering estimation algorithm containing parameter tracking is adopted to carry out joint estimation on the rotating speed of the induction motor and the rotor flux linkage; high-performance estimation of the rotating speed of the induction motor and the rotor flux linkage can be effectively realized; compare with EKF, the STF algorithm including parameter tracking is more excellent in estimation precision, tracking speed and stability; moreover, the abrupt change state can be quickly tracked, and particularly, relatively good estimation performance can still be kept in a low-speed section; the state estimation precision and the algorithm robustness are effectively improved; and a foundation is laid for realizing speed-sensor-free high-performancevector control of the induction motor.

A system and method for generating dynamic user interfaces that retrieve, at a backend UI server, one or more user requests originating from an information management application, retrieving, at the backend UI server, a data response originating from a data server in response to the one or more user requests, calculate, at the backend UI server, relevance values for each display component of the information management application, append the calculated relevance values to the data response, transmit the calculated relevance values and the data response to a frontend UI server; and generate, at the frontend UI server, a user interface for the information management application based on the calculated relevance values.

The invention discloses a class-schedule social method and a system thereof. A central server is responsible for centralizedly managing registration and login of all the types of account numbers. Through inputting a teacher account, teaching information can be acquired, communication connection with a student account and a parent account can be established, a student and a parent can be contacted at any time, online work assignment can be performed and notification omission can be avoided. Through inputting the student account, all the participated courses in this period can be acquired. Through clicking a course link, course prompt preparation or a course summary can be consulted, which is convenient for the student to carry out preview and review work. Instant communication can be directly established with the teacher so that questions about the course can be answered. Through clicking a name of the student in a class schedule, instant communication with the student is established, and a learning discuss group is conveniently established so that student social initiative and learning efficiency are increased. Through inputting the parent account, course information and teacher information of a corresponding student account can be acquired so that the parent can know learning dynamic of the student.

The invention provides a multidimensional learning ability improving system. According to the invention, a user terminal includes a test unit, a test result output unit and a habit nurturing unit, wherein the test unit is for a user to complete a test, the test result output unit outputs the test result of the test unit and transmits the test result to a management terminal, the habit nurturing unit, based on a study plan, generates an alert message so as to remind the user to study, the management terminal includes a plan generation unit which generates a study plan on the basis of the test result of the user, and transmits the study plan to the corresponding user terminal. The system herein can stage tests for students, learn students' intelligence levels, study habits, characters, interests, competencies and values on the basis of the test results, learn students' advantages, generate corresponding study plans, plan students' study purpose through study plans, activate studying potential in a multidimensional manner, help students to nurture good study habits by means of the habit nurturing unit, help students to take initiative to study, and increase students' learning ability in a comprehensive manner.

The invention discloses a game theory optimization method for carrying out distributed solution on a sudoku problem and gives physical game realization of the sudoku problem. The method comprises the following steps: (1), building a utility function and proving that the sudoku problem can be converted into a potential game model; (2) optimizing the states of participants step by step by using learning motivation so that the states of the participants reach the optimal states, namely Nash equilibrium point of the potential game.

The invention discloses a statistical data-based mobile interactive training evaluation method and system for marketing services. The method includes the following steps: 1) establishing an evaluation index system for the student's learning enthusiasm level; 2) evaluating the student's learning enthusiasm by using the analytic hierarchy process . The system includes: a system establishment module, which is used to establish an evaluation index system of the student's learning enthusiasm level; an evaluation module, which is used to evaluate the student's learning enthusiasm by using the analytic hierarchy process. The present invention firstly establishes the evaluation index system of the student's learning enthusiasm level, and then uses the analytic hierarchy process to evaluate the student's learning enthusiasm, which can timely and efficiently grasp the student's learning dynamics, stimulate the enthusiasm of "teaching" and "learning", and improve the deficiencies of the system module functions , Comprehensively improved the level of interactive training teaching ability.

The invention discloses an online teaching and learning attention detection technology, which comprises a camera, an information processing module, a database, a first analysis module, a second analysis module, a data integration module, a controller, a display, a memory and a portable device. The camera is used for real-time monitoring the learning dynamics of the students and transmitting the learning state video to the information processing module. The invention obtains the learning state video of the student through the camera, then the video is processed by the information processing module, when the student appears distraction, the first analysis module and the second analysis module can analyze the student's line of sight and the ear part, and detect whether the student's eyes arewatching elsewhere or sleeping with his head bent down, or whether the student has headphones to listen to audio information irrelevant to the learning video; and when the student appears distraction,the first analysis module and the second analysis module can analyze the student's line of sight and the ear part. Through the two real-time monitoring modules whether students are seriously studyingcan be monitored; the monitor can show whether the current student is distracted or not in real time.

The invention relates to a graph-based earth surface solar radiance prediction method fusing space-time attention. The method comprises the following steps: selecting SWR sub-grids covering actual measurement sites to construct a graph, meteorological grid points and ground radiance actual measurement sites as nodes of the graph; representing the graph structure and the time information by using space embedding and time embedding; learning dynamic relationship associativity among nodes in the graph by using an attention mechanism fused with time and space; adding an attention transformation module between an encoder and a decoder, and modeling a historical and future relation through space-time embedding; and adopting the node characteristics of the actual measurement site as the input ofa decoder to predict the solar radiance. According to the method, the spatial-temporal correlation is fully considered, and the radiance can be predicted more accurately.

The invention discloses a behavior recognition method and device based on deep learning, equipment and a storage medium. The method comprises the steps of collecting walking behavior information of pedestrians on a preset road section within a preset time range before a red light is turned on; extracting current static image information and current optical flow image information in the walking behavior information; performing behavior recognition on the current static image information through a deep learning static image behavior recognition model to obtain a static behavior recognition result; performing behavior recognition on the current optical flow image information through a deep learning dynamic image behavior recognition model to obtain a dynamic behavior recognition result; performing behavior recognition on the walking behavior information according to the static behavior recognition result and the dynamic behavior recognition result; and performing behavior recognition on the walking behavior information according to the static behavior recognition result and the dynamic behavior recognition result, thereby improving the behavior recognition accuracy.

The invention discloses a power grid dispatching knowledge graph data optimization method and system. The method comprises the following steps: firstly carrying out automatic mining on high-quality phrases of a field through a deep learning method, and completing automatic recognition and equivalent disambiguation of a dispatched entity; then, completing dispatched entity global relationship extraction according to deep learning technology, so as to complete entity relationship recognition and verification, and achieve the purpose of establishing an initial power grid dispatching knowledge graph; on the basis that the two steps are completed, using a natural language learning knowledge fusion technology, and conducting incremental training on newly-added scheduling plan data based on timestamps; meanwhile, introducing life cycle management of knowledge graph knowledge content in the completion process of the steps; finally, completing a continuous learning dynamic knowledge graph under the cooperation of the steps. According to the invention, the high precision of a power grid dispatching optimization decision knowledge graph is ensured, and the consumption of computing resources and time during updating training is reduced while dynamic updating of incremental knowledge is ensured.

The invention discloses an artificial intelligence olfactory dynamic response spectrum gas detection and recognition method. The method comprises the following steps: collecting leakage gas data through dynamic signals of the array sensors; standard dynamic response spectrum reconstruction is carried out; establishing a standardized data matrix and a vector spectrum for to-be-detected gas collected by a sensor array, performing feature extraction and training learning on picture data in a standard spectrum library, establishing a machine learning dynamic response spectrum recognition model, and performing quantitative and qualitative recognition on the gas by utilizing the machine learning spectrum recognition model. According to the invention, traditional single-sensor response identification is converted into a multi-dimensional sensor dynamic response spectrum; Gas detection and recognition are achieved through an automatic map recognition method, the defects of singleness and crossinterference of a traditional single sensor in the aspect of gas detection are overcome, different gases are rapidly and accurately detected through the same sensor array, the detection efficiency and precision are improved, and meanwhile the detection result is visualized and more visual.