48 results about "Hybrid learning" patented technology

In a network that executes a link state routing protocol, a network node receives periodic disseminations of link state information from other network nodes. The link state information includes neighboring node identity and link cost metrics. The network node calculates the initial routing paths based on the received link state information by using a link state routing algorithm. It then adapts the calculated path based on both the current link state information and past link state information through a reinforcement learning process. The network node then selects a routing path to each destination node based on the adaptation and updates the routing table accordingly.

Aspects of the present invention include object detection training systems and methods and using object detection systems and methods that have been trained. Embodiments presented herein include hybrid learning approaches that combine global classification and local adaptations, which automatically adjust model complexity according to data distribution. Embodiments of the present invention automatically determine model complexity of the local learning algorithm according to the distribution of ambiguous samples. And, embodiments of the local adaptation from global classifier avoid the common under-training problem for local classifier.

A hybrid learning environment may include a hybrid learning environment system; a learning center, including a first client computer, and an online class display coupled to the first client computer to display instructional material; a second client computer at a remote location displaced from the learning center to present instructional material from the hybrid learning environment system server; a database accessible to the hybrid learning environment system to store at least one of the instructional material; and at least one network coupling said first client computer, said second client computer and said hybrid learning environment system server.

The invention discloses an automatic complicated target identification method based on a hierarchical object semantic graph, and relates to a target identification technology. The automatic complicated target identification method comprises the following steps of: establishing a multi-class complicated target image representative set; performing multi-scale partitioning on an image of a training set, gradually calculating characteristic information of each part object, and constructing a hierarchical semantic graph; counting partial characteristic attributes of objects by using a judgment type classifier by adopting a spiral mixed learning mode, calculating interactive influence among the objects by combining a generation type message transmission mechanism, and deducing and calculating the hierarchical semantic graph; and resolving a target of interest in the image by using the hierarchical object semantic graph obtained by learning, and realizing positioning, extraction and type identification of a plurality of classes of complicated targets. The method is relatively high in intelligentization degree; and requirements for identifying a plurality of classes of complicated targetsin natural and remotely sensed scene images and explaining the images can be met.

The invention disclosed relates to a learning system and related methods for providing training and continuing education to members of an organization's supply-chain workforce. A learning system utilizes a blended learning approach that matches various educational delivery techniques to the appropriate learning materials, time constraints, and skill level of the end user. The system is simultaneously accessible to multiple organizations in order to provide uniformity of learning among a wide variety of participants within the supply-chain workforce, while remaining adaptable to the unique learning needs of each organization. Thus, the invention captures the benefits of distance learning techniques while minimizing the isolationist learning environment typically associated with those leaning delivery methods. Furthermore, the invention seeks to capture the benefits associated with traditional learning delivery methods such as team-oriented learning and personal instructor feedback.

The invention discloses a mixed learning model-based DSP real-time bidding advertisement system. The system comprises an RTB advertisement bidding engine module, an offline model training module and an online model training module, wherein the RTB advertisement bidding engine module is connected with the offline model training module and the online model training module, calls a CTR prediction engine to predict a CTR value of a corresponding advertisement in combination with a CRT model generated by one or two of the offline model training module and the online model training module, and performs corresponding advertisement release.

An online Gaussian mixture learning model for dynamic data utilizes an adaptive learning rate schedule to achieve fast convergence while maintaining adaptability of the model after convergence. Experimental results show an unexpectedly dramatic improvement in modeling accuracy using an adaptive learning schedule.

The invention relates to a Domain-floux botnet detection method based on hybrid learning. The method comprises the following steps: distinguishing a training data set and a detection data set from input DNS data, respectively preprocessing the training data set and the detection data set, inputting the preprocessed training data set into a model, training to obtain a classifier, inputting the preprocessed detection data set into the classifier, outputting clustered domain name clusters, calculating domain name cluster scores, and screening domain name cluster scores which belong to Domain-fluxbotnet domain name; and obtaining the IP address of the infected host and the IP address of the C & C server. According to the invention, a foundation is laid for subsequent defense measures, and characteristics related to domain names, time, request IPs, resolution IPs and the like are used during resolution and are not easy to bypass; the advantages of supervised learning and unsupervised learning are considered, the classification and clustering algorithms are combined, unknown zombie samples which are greatly different from a training set and different in expression form can be effectively detected, and the detection speed is higher than that of other clustering algorithms.

The invention provides a training method and using method of a voice wake-up hybrid model and related equipment; during model training, a voice wake-up hybrid model is obtained through hybrid learningtraining of a voice separation network, a feature transformation network and a wake-up word detection network. When the model is used, an audio to be recognized is input into the voice wake-up hybridmodel, and the wake-up probability is directly obtained. When the wake-up probability is greater than the threshold, the wake-up word identified from a to-be-identified audio is judged. In a model training process, a first loss function obtained through the voice separation network and a second loss function obtained through the wake-up word detection network are weighted to obtain a comprehensive loss function, and weight parameters of the voice separation network, the feature transformation network and the wake-up word detection network are obtained through back propagation and learning according to the comprehensive loss function. Meanwhile, the networks are unified into one framework, and a joint optimization method is used, so that the model can learn optimal separation and wake-up network parameters at the same time, and the wake-up rate is effectively improved.

The invention discloses a high-order neighborhood hybrid network representation learning method and a device. A self-attention mechanism and a cascade aggregation layer are added on the basis of an original graph convolution layer, and the method comprises the following steps: converting a Laplace matrix of a graph into a node pair graph attention matrix by using the self-attention mechanism, andtraining weight parameters to learn different attention coefficients; gathering information flows with different distances through the cascade aggregation layer, and taking the output of the previousorder as the input of the next order to control the calculation complexity; and determining that the embedded vector is output to a downstream machine learning task, or outputting a classification result. According to the method, real end-to-end training can be realized, the training speed of the model is effectively improved, and the proposed idea of network high-order and low-order information hybrid learning has field expandability and is simple and easy to realize.

The invention belongs to the field of behavior recognition, particularly relates to a behavior feature extraction method, system based on space-time frequency domain hybrid learning, and a device, andaims to solve the problem of low skeleton behavior feature extraction precision. The method comprises the steps of obtaining a video behavior sequence based on a skeleton, and extracting a time-spacedomain behavior feature map through converting a network; inputting the time-space domain behavior feature map into a frequency domain attention network, performing frequency selection, inverting toa time-space domain, and adding the obtained behavior feature map to the time-space domain behavior feature map; synchronously performing local and non-local reasoning, and performing high-level localreasoning; and globally pooling the time-space domain behavior feature map obtained through reasoning to obtain the behavior feature vector of the video behavior sequence. The method can be applied to behavior classification, behavior detection and the like. According to the method, an effective frequency mode is adaptively selected in a frequency domain, a network with local affinity fields andnon-local affinity fields is adopted in a time-space domain for space-time reasoning, local details and non-local semantic information can be synchronously mined, and therefore the behavior recognition precision is effectively improved.

The invention discloses a WSN wireless communication module fault diagnosis method based on a fuzzy neural network. A fuzzy neural network current model is established by using emission consumption parameters corresponding to a DHT11 temperature and humidity sensor under different temperatures and voltages for the fault diagnosis of a wireless communication module. For data subjected to normalization processing, firstly an initial structure and parameters of the fuzzy neural network are adaptively determined by using subtraction clustering, then parameter optimization and adjustment are carried out on the model by using a hybrid learning method combining the particle swarm optimization algorithm with the least square method, and finally fault diagnosis is carried out on a test sample by using a trained diagnosis model. According to the WSN wireless communication module fault diagnosis method disclosed by the invention, the advantages of fuzzy reasoning and the neural network are integrated, an improved learning algorithm is adopted, the fuzzy neural network current model of the wireless communication module is established for the relation among the current, the voltage and the faults of a WSN, and the model is short in training time, high in convergence speed and high in fault diagnosis efficiency.

The invention relates to the technical field of computational intelligence, in particular to an interval type II intuitionistic fuzzy random vector function connection neural network design method. The interval type II intuitionistic fuzzy random vector function connection neural network design method comprises the following steps: constructing an interval type II intuitionistic fuzzy random vector function connection neural network; adopting a hybrid learning method combining quantum clustering algorithm, swarm intelligence algorithm and least square method to adjust the structure and parameters of the neural network. The neural network adopts the membership function and the non-membership function based on the beta function as the forepart of the fuzzy rule, and adopts the random vectorfunction to connect the neural network as the afterpart of the interval fuzzy rule. The IT2IF-RVFLNN designed by the invention can achieve global approximation.

The present invention relates to a method for managing hybrid learning, a management device for hybrid learning, a computer readable storage medium, a student terminal for hybrid learning, a teacher terminal for hybrid learning, and a system for hybrid learning. In order to combine the advantages of online courses and offline courses and make up the defects of the online courses and the offline courses to a certain extent, the invention provides hybrid learning, namely learning combining the online courses and the offline courses. The method provided by the invention comprises the following steps: publishing online courses and jobs, tests and / or questionnaires corresponding to the online courses; Collecting feedback data input by the student in the homework, the test and / or the questionnaire after finishing the online course; And generating offline curriculum schedule by analyzing the feedback data. According to the hybrid learning, the learning process can be optimized, the learning efficiency can be improved, and the learning quality can be ensured.

The invention discloses a network micro-course system based on cloud data, which comprises a client, a teacher terminal and a cloud data center, and is characterized in that the cloud data center is respectively in network connection with the client and a data acquisition terminal; The client is used for sending a lecture attending request to the cloud data center, feeding back suggestions, receiving and displaying the lecture video sent by the cloud data center, and carrying out teaching interaction with the teacher terminal through the cloud data center; The teacher terminal is used for carrying out micro-course construction, uploading related data to the cloud data center for storage, and carrying out teaching interaction with the client through the cloud data center; The cloud data center comprises a cloud database, an identity verification unit, a micro-course directory generation unit, a teaching interaction unit, a data processing unit and a data transmission unit. According tothe invention, micro-courses are created and teaching application is carried out, and collaborative learning, hybrid learning and cloud computing auxiliary teaching are closely combined, so that the micro-courses play a more powerful role, and a teaching mode of the courses is favorably changed.

The invention discloses a method and a system for predicting a vehicle braking intention based on a hybrid learning mode, which belong to the field of predicting the vehicle braking intention and can perform braking control and management of braking energy in advance so as to improve the braking performance of a vehicle and improve the energy recovery efficiency. The method comprises the following steps of: collecting data and performing smoothing processing and normalization processing on the collected data; performing division of the braking intentions into slight braking, common braking, strong braking and emergency braking based on an unsupervised learning mode fuzzy C-means algorithm; performing selection of the feature parameters based on a feature selection algorithm ReliefF; and finally, performing offline training of a braking intention prediction model based on a supervised learning algorithm Adaboost. The system mainly comprises: a pressure sensor, a motor controller, a battery controller, a transmission controller, a CAN bus, a peripheral circuit and an electronic control unit (ECU). The prediction model obtained through offline training is located in the ECU for online prediction.

The present invention relates to the design and development of a learning game platform, system and method that is implemented in an electronic device to promote autonomous work and active learning, integrating all students, with self and peer assessment, supported by instructional videos that can be used for any subject / course implemented on electronic devices for the teaching and learning of mathematics, languages, physics or any other subject. Students can use this in the classroom or outside the classroom in an online, blended or flipped learning setup to study and solve problems, combining analog input and digital outputs.The innovative part of this platform is that it integrates: analog input; a gamification scheme; videos; levels of problems and levels of instructional videos, organized in a video-book; digital-portfolio; self-assessment and peer-assessment.Different difficulty levels of problems and of video instructional resolutions, self and peer-assessment is implemented to include all students and promote autonomous and active learning in a gamified platform to motivate students.The application of the present invention can be used for online, mobile, blended and flipped learning. Teachers can follow what students are doing in another back office application, which is also part of the system of the present invention.The present invention can be implemented also on electronic devices such as smartphones, tablets, etc. allowing students to use this it in the classroom or outside the classroom in a blended learning model to solve problems.The learning gamified platform, system and method, for an electronic device executing an application, comprises a server for a reading medium and storing user information, scores and user progress and e-portfolio within a play and learn app platform. The server communicates through the Internet with electronic devices having Internet access operated by at least one player. Registration means allows the player to register with and log on to the play and learn app. Also the app allows the teachers to follow the performance of the players and to improve some aspects of the teaching program in a given classroom.

A hybrid learning system is provided for searching an experimental space. A data mart is configured to acquire, store, and manipulate a set or meta-set of data including at least historical experimental data, descriptor data, and concurrent experimental data. A search engine is designed to use unsupervised learning techniques to select a set of evaluation points representing a corresponding set of experiments to be run, based on data from the data mart. A point evaluation mechanism provided with supervised learning modules which perform predictive processing based on the evaluation points selected by the search engine, and a scoring module performs a rating operation on outputs of the learning modules to rate the outputs of the learning modules from best to worst. The data mart search engine and point evaluation mechanism allow for a repetitive processing to refine an output of potential solutions without the requirement of continually running actual physical experiments.

The invention discloses a novel hybrid learning environment teaching platform, relating to the technical field of teaching platforms. The platform comprises an equipment group and a server group whichare both installed in a traditional classroom. Teachers and students are in communication connection with the server group through network equipment, and the traditional classroom and a MOOC classroom jointly carry out teaching and learning resource information interaction through the server group. A set of synchronous teaching platform integrating theories, experiments and innovative practices is established by utilizing a communication interconnection function of a campus network. The four-dimensional integration three-class fusion integrated college classroom teaching function which integrates three teaching links of theories, experiments and innovative practices is achieved. Through a unified browser interface of the platform, a teacher can start experiment and innovative practice equipment of a remote laboratory in a classroom through a campus network, integrated synchronization of three teaching links is achieved, and students can complete three key links of teaching in the classroom on the same interface at the same time.

The invention relates to a new sentiment classification model based on hybrid learning. In the first stage, an improved dictionary classification method is used for calculating emotion scores on a whole data set, and data with extremely high scores or extremely low scores are directly marked; in the second stage, the rest is used to calculate the emotion score based on the emotion dictionary and the BI-GRU fusion model, and the two-stage hybrid framework enables the method to be effectively applied to emotion classification. Experiments show that a single model is unsatisfactory in emotion classification effect under various complex contexts, high in difficulty and low in precision, and error preference of the single model can be effectively improved by adopting a multi-model fusion method, so that the classification effect is improved.

The invention provides a fuzzy neural network algorithm integrating classification and clustering into one body. Each step is redesigned based on the fuzzy minimum-maximum neural network. In each step, the possibility of simultaneous existence of unidentified and identified samples is fully considered to distinguish the difference between classification learning and clustering learning, so that two learning methods coexist in the same fuzzy neural network system. If an input sample set is composed of all unidentified samples, an unsupervised learning mode is used to cluster the sample set. If the input sample set is composed of all identified samples or some identified samples and some unidentified samples, a supervised learning mode is used to classify the sample set. The algorithm provided by the invention can be used for pure clustering and pure classification, and can be used for a hybrid learning mode of clustering and classification. In the process of hybrid sample learning, both identified and unidentified samples are fully utilized, which improves the classification accuracy.

The invention discloses a wireless federated learning method and device, and the method comprises the steps: enabling a distributed learning user based on air computing and a centralized learning user based on non-orthogonal multiple access to share uplink spectrum resources through the concurrent transmission assisted by a dual-function intelligent metasurface, the local data and the model parameters are sent to a base station at the same time for mixed learning; a base station carries out signal decoding on centralized users based on non-orthogonal multiple access so as to obtain local data of each centralized learning user, and the local data are used for calculating model parameters of the centralized learning users; then the base station obtains average model parameters of federated learning users by using serial interference cancellation and air computing technologies; the base station updates a global model by combining the two types of model parameters; and then the base station issues the global model to all federated learning users to carry out the next round of learning until the global model converges or reaches the maximum number of iterations. According to the method, the communication overhead and the transmission delay can be remarkably reduced, and better learning performance can be obtained.

The invention provides a cross-domain pedestrian re-identification method and system based on multi-feature mixed learning, and belongs to the technical field of computer vision. The method comprises: by means of a re-identification model subjected to combined training, extracting pedestrian global features, pedestrian attribute features and pedestrian local features of the pedestrian image to be recognized and a bottom library image, which is similar to the pedestrian identity in the pedestrian image to be recognized, in the image bottom library gallley; and fusing the extracted to-be-identified features, and carrying out similarity matching sorting on the fused features of the features of the bottom library image to obtain a pedestrian re-identification result. According to the method, inter-domain joint training and multi-feature mixed learning are utilized to reduce inter-domain differences, so that the system is more stable and higher in robustness, source domain training of global and local features and joint training of attribute features are performed on images of different scenes, pedestrian attributes are combined, the adaptive capacity of a cross-domain pedestrian re-identification model is improved, and pedestrian re-identification is carried out on a cross-domain data set, so that the cross-domain pedestrian re-identification performance is improved.

The invention relates to a population spatialization method based on AWA-DRCN, and belongs to the technical field of Internet and computers. The method comprises the following steps: S1, acquiring a population data set and driving factor data, and preprocessing the population data set and the driving factor data; s2, a de-coarsening process: adopting an AWA model to achieve global feature learning, and completing regional and county-level population general survey data rasterization; and S3, a'refining'process: constructing a high-efficiency sub-pixel convolutional neural network DRCN model, taking the rasterized population data and the driving factor data in the step S2 as features, taking the processed grid population raster data as labels, inputting the processed grid population rasterdata into the DRCN model for hybrid learning to achieve global and local features, and obtaining a population spatialization result in spatial distribution. The AWA + DRCN model adopted in the methodis high in simulation precision, and the space residual value interval error is minimum.

The invention relates to a driver steering intention prediction method based on hybrid learning, and the method is characterized in that the method comprises the following steps: 1) carrying out the multi-mode data collection on a driving simulation platform, and carrying out the preprocessing of the collected multi-mode data; 2) establishing a hybrid learning time sequence model based on the preprocessed multi-mode data; and 3) loading a hybrid learning time sequence model on the intelligent vehicle, and inputting the driver electromyographic signal sequence acquired on line into the hybrid learning time sequence model for prediction to obtain a driver continuous steering intention prediction result and a discrete steering intention prediction result. By establishing the hybrid learning time sequence model, continuous steering torque prediction and discrete intention classification prediction are achieved, and accurate prediction of the steering intention within a certain prediction time range can be realized by setting historical observation parameters. The method can be widely applied to the field of driver steering intention prediction.