42 results about "Local learning" patented technology

Method, comprising: receiving an indication of one or more supported bit-widths for local learning by a first node among plural nodes; generating a respective quantized version of a model for at least one of the supported bit-widths; providing the generated respective quantized versions of the model for the at least one of the supported bit-widths or a link to location from where the first node may download the at least one quantized version of the model for the at least one of the supported bit-widths to the first node.

Described is a Neuromorphic Adaptive Core (NeurACore) cognitive signal processor (CSP). The NeurACore CSP includes a NeurACore local learning layer block that is operable for receiving as an input a mixture of in-phase and quadrature (I / Q) signals and mapping the I / Q signals onto a neural network to determine complex-valued output weights of neural states of the neural network. A global learning layer is included that is operable for adapting the complex-valued output weights to predict a most likely next value of the input I / Q signal. Further, a neural combiner is included that operable for combining a set of delayed neural state vectors with weights of the global learning layer to compute an output signal, the output signal being separate in-phase and quadrature signals.

The invention discloses a remainder value evaluation method and device for second-hand electronic equipment and a medium, belongs to the technical field of electronic equipment value evaluation, and aims to solve the problems that in existing remainder value evaluation of the second-hand electronic equipment, manual detection evaluation is generally adopted, the subjectivity is high, the remainder value detection dimension is shallow, the labor time consumption cost is high, and the evaluation efficiency is high. And dynamic data support is also lacked. The method comprises the following steps: carrying out structural data acquisition processing related to a physical appearance mode on second-hand electronic equipment to be evaluated; performing data test processing based on a hardware performance mode under non-invasive flaw detection on the second-hand electronic equipment; using mode quantitative analysis under federal local learning is carried out on the software system using mode of the second-hand electronic equipment; carrying out equipment use mode scoring processing under the microscopic state analysis network; and carrying out residual value evaluation processing in a related market residual value evaluation interval between the comprehensive state score and the multi-dimensional value influence factors of the second-hand electronic equipment.

A model generation apparatus according to one or more embodiments may include: a generating unit that generates data using a generation model; a transmitting unit that transmits the generated data to a plurality of trained identification models that each have acquired, by machine learning using local learning data, a capability of identifying whether given data is the local learning data, and causes the identification models to perform an identification on the data; a receiving unit that receives results of identification with respect to the transmitted data executed by the identification models; and a learning processing unit that trains the generation model to generate data that causes identification performance of at least one of the plurality of identification models to be degraded, by performing machine learning using the received results of identification.

The federated learning method is performed by sampling at least one edge device to perform federated learning based on local data of each edge device, performing local learning on the sampled edge device to aggregate the results of the local learning, generating global weights based on the aggregated results of the local learning, and broadcasting them to each edge device.

The invention provides a channel gain map reconstruction method based on a k-nearest neighbor refined neural network, relates to the technical field of wireless communication, and has the technical key points that the invention provides a two-stage CGM reconstruction framework under a sparse regular grid sampling condition, and in the stage 2, local learning is carried out during query to refine prediction, so that explicit decoupling and robust extrapolation are realized. K nearest neighbors are selected from a total sample by adopting a global k nearest neighbor thought which is not constrained by a clustering boundary, spatial continuity is preferentially utilized, and k nearest neighbors in a cluster which can be used as a contrast are selected only in the same cluster. The whole process does not depend on LoS / NLoS labels or high-cost environment prior, the CGM can be reconstructed only by regular grid sampling measurement, and high-precision benefits are kept under different sampling interval values d and k. On the premise of not increasing expensive measurement and label cost, the MSE of CGM reconstruction is obviously reduced.

The application discloses a vehicle image recognition system based on local feature learning, and relates to the field of vehicle image recognition.The system comprises an image preprocessing module, which is used for denoising an input image; an image segmentation module, which is used for segmenting the image into a plurality of superpixel regions; a dense feature extraction module, which is used for extracting local gradient features of the image and constructing a statistical histogram; a feature recombination and representation module, which is used for clustering analysis on the local gradient features and reconstructing feature distribution into a category distribution histogram; a local learning model construction module, which is used for selecting representative samples as a local learning center and constructing a local training subset; and a collaborative classification module, which comprises a local classifier, is used for training respectively and selecting a local classifier for category discrimination, and finally determines a vehicle region.The application scheme can realize high-accuracy and high-stability vehicle target recognition in a static picture without relying on time sequence information.

The application discloses a high-efficiency local learning system supporting an arbitrary network topology, which is applied to various local learning tasks and is realized by an electronic device with computing capability and comprises the following modules: an rlayer module, a datasets module, a grads module, a metrics module and an optimizer module; operator classes, gradient solving classes, optimizer classes, dataset classes and measurement index functions in an operator layer are used for providing interfaces of the modules in the module layer; the operator classes, the dataset download and preprocessing classes, the optimizer classes, the gradient derivation classes and the index calculation functions in the module layer all have input and output interfaces, a user can create and instantiate a network on demand through the rlayer module, download and preprocess a dataset through the datasets module, select a corresponding optimizer through the optimizer module, select a local learning gradient derivation class through the grads module, select a loss and precision calculation function through the metrics module, complete initialization of the network and corresponding hyperparameters, and realize network training and testing.

This invention discloses a method, device, and medium for assessing the residual value of used electronic devices, belonging to the field of electronic device valuation technology. It addresses the technical problems of existing used electronic device residual value assessment methods, which generally rely on manual inspection and evaluation, are highly subjective, have shallow residual value detection dimensions, are time-consuming and costly, and lack dynamic data support. The method includes: structured data acquisition and processing of the physical appearance modalities of the used electronic device to be assessed; data testing and processing of the hardware performance modalities of the used electronic device based on non-invasive flaw detection; quantitative analysis of the usage patterns of the software system usage modalities of the used electronic device under federated local learning; and equipment usage modal scoring processing under a micro-state analysis network; and residual value assessment processing within a relevant market residual value assessment range based on the comprehensive state score and the multi-dimensional value influencing factors of the used electronic device.

The invention discloses a block chain-based federated learning cross-border trade service platform model verification method, and belongs to the technical field of block chain federated learning of cross-border trade, and the method comprises the steps: in the federated learning of the current round, each client node carries out the local learning of a global model through employing a first data set, and obtains a local model, submitting the data to all verifier nodes; each verifier node trains the global model by using the second data set to obtain a verification model, verifies the quality of each target local model based on the verification model to obtain each verification result, and submits each verification result and the corresponding target local model to an aggregation node; and the aggregation node aggregates each target local model based on all verification results to obtain a new global model, and adds the new global model to the block chain network. According to the method, the quality of each local model can be accurately verified, so that the quality of the aggregated global model is guaranteed.

The invention discloses a federal learning model aggregation optimization method for adaptive adjustment of weight and learning rate, and the core of the method lies in collaborative optimization of aggregation weight and local learning rate: on one hand, through introducing an adaptive weight mechanism, the contribution degree of a low-performance client is improved, and the low-performance client is prevented from being marginalized in long-term training; on the other hand, an appropriate learning rate is configured for the client according to the computing power, the learning process of the vulnerable party is accelerated, and global convergence is balanced. According to the method, the problems of slow convergence and generalization reduction caused by calculation performance difference of clients in federated learning are solved, the method has remarkable advantages in convergence speed, communication round and final precision, and the training efficiency and accuracy are improved.

Provided is a federated learning model, a generation apparatus thereof, and the like that easily and suitably contribute to marketing activities. A federated learning model generation apparatus includes a local learning model acquisition unit and a federated learning model generation unit. The local learning model acquisition unit acquires a plurality of different local learning models that have learned a relationship between a plurality of customer groups respectively generated from business customer data owned by a plurality of business operators and consumption behaviors corresponding to the business operators. The federated learning model generation unit receives a predetermined consumption behavior of a customer as input data by federating at least a part of the acquired local learning models, and generates a federated learning model that outputs prospective customer data for the input data.

The present invention relates to a learning system, a learning method, and a learning computer program product. Each local learning device (2) of the learning system (1) learns a subset of correction weight coefficients that corrects a part of a set of weight coefficients of a base model that is a basis of a generative model using a local training dataset, generates distribution data that represents a distribution of features of data included in the local training dataset, and transmits the learned subset and the distribution data to a server (3). The server (3) generates a simulation training dataset that reproduces a distribution of features represented by the distribution data based on the distribution data received from each local learning device (2), and learns a gate network used for selecting a subset of correction weight coefficients to be used using the simulation training dataset.

The application provides a credit risk control prediction model training method, a credit risk control prediction method and a system, and relates to the technical field of artificial intelligence. The training method comprises the following steps: inputting the data in the respective training sets of the two credit institutions into respective local learning models, obtaining the hidden features of the data in the respective training sets, and using the hidden features to complete the data in the training sets; inputting the data into the local prediction model; comparing the model output result with the corresponding risk label to obtain a loss value and an alignment loss value; sending the loss value of one party to the other party to obtain a reconstruction loss value and a comprehensive gradient, which is equivalent to using the user data of the self and the user data of the other party to train the local prediction model, thereby avoiding the data island problem; updating the parameters of the local prediction model according to the comprehensive gradient until the loss value and the reconstruction loss value of each party converge to the corresponding preset accuracy, and stopping the training. The local prediction model is used as a credit risk control prediction model to ensure the safety of loan funds.

The application relates to an activation learning method, system and picture classification method of an artificial neural network for picture classification, which comprises the following steps: constructing an artificial neural network while taking data and category labels or multi-modalities as inputs; training the artificial neural network through an unsupervised local learning method, so that the output activation intensity of the last layer of neurons in the artificial neural network reflects the typicality of input samples; inputting data to be identified into the trained artificial neural network and optimizing missing category labels so that the output activation intensity is maximum, thereby obtaining a data classification result; or inputting category labels into the trained artificial neural network with random noise and optimizing missing data parts so that the output activation intensity is maximum, thereby realizing data generation. The application can better learn the statistical probability distribution of input sample data, unifies the framework of supervised learning, unsupervised learning and a generation model, and has strong practicability.

The application provides a personalized federated learning method and system of a shared model, comprising: a server end saving and initializing a shared model, a global model and a global class prototype of a client, and the client initializing a local learning weight vector; the server sending a shared model of other clients to one client, the client setting the local shared model as the global model and training, and uploading the shared model and the local class prototype to the server end; and the server calculating the global class prototype and the global model according to all received shared models and local class prototypes. The application solves the knowledge sharing problem in federated learning, improves the performance of the personalized model, solves the deviation problem in the local model training of the client, and the obtained personalized prototype contains more global information than the personalized prototype, so that the effective fusion of the global knowledge and the local characteristics is realized while reducing the state dependence and the communication overhead, so as to improve the robustness and adaptability of the model in a complex scene.

One aspect of the present invention relates to an inference apparatus that obtains inference results for each inference model by providing object data to multiple inference models derived from local learning data obtained in different environments to perform prescribed inferences, and by having each inference model perform prescribed inferences. Then, the inference apparatus generates an inference result for the object environment by determining the values ​​of each integration parameter based on the environmental data, weighting the inference results of each inference model using the determined values ​​of each integration parameter, and integrating the weighted inference results.

The invention discloses an online training cerebellar neural network controller and a control method for a single-inductor multi-output switching power supply. The method comprises the following steps: performing difference operation on reference voltage and output voltage through a logical operation unit to obtain an output voltage error, an error integral quantity and an error differential quantity; each PID controller is configured to enable an input signal to correspond to each path of output voltage error, and is used for carrying out proportional-integral-differential adjustment on the PID controller based on each error integral quantity and error differential quantity; the cerebellar neural network is configured to update the weight of the cerebellar neural network in real time through online training based on a local learning rule and output a digital control signal carrying duty ratio adjustment information; the digital pulse modulator generates a complementary switching signal based on the duty ratio adjustment information and forms a switching signal to drive a power switching tube of the single-inductor multi-output switching power supply, the disturbance recovery time and overshoot of the digital power supply can be effectively reduced, and the transient performance is remarkably improved.

The application discloses a hyperspectral image multi-class change detection method based on global-local joint modeling, belongs to the technical field of remote sensing image processing, and solves the problems of calculation redundancy and receptive field limitation of a local learning method and insufficient modeling capability of a global learning method in the prior art.The method comprises the following steps: constructing a hyperspectral image change detection model; inputting a joint hyperspectral image to be detected into the hyperspectral image change detection model; obtaining change class prediction results of each pixel; and generating a hyperspectral image multi-class change detection map.The embodiment of the application constructs a hyperspectral image change detection model, utilizes the linear calculation complexity advantage of a Mamba model, efficiently establishes long-range dependence of spatial information on the whole image, realizes fast reasoning without dividing image blocks, enhances the capturing capability of the model for subtle change regions and edges, and significantly improves the precision and robustness of hyperspectral image multi-class change detection.

The present application relates to shallow geothermal energy development and utilization technology, especially to a kind of energy underground subway station intelligent operation and maintenance method, with the federal learning operation and maintenance control model constituted by central server-local client, the various local learning model training of 5 modules is carried out on local client, the calculation result is encrypted and uploaded to central processor by global federal learning model, the result obtained by global model is again issued to the local model of each local client, through the repeated iteration between client-central server, until global model reaches stability, finally make different module regulation and control scheme matched with each other.The present application can be used for real-time regulation and early warning;Realize the integrated regulation and control between station user end, unit equipment, energy tunnel, energy bottom plate, energy support pile multiple modules, the control scheme obtained is scientific and reasonable, improve the stability and service life of shallow geothermal energy underground subway station environmental control system.

The invention discloses a knowledge tracking method and system based on dynamic local sampling and multi-scale memory transfer, and the method employs an innovative multi-level architecture, and comprehensively captures the dynamic evolution process of the knowledge state of students through the organic combination of deformable sampling, multi-scale feature fusion and short-term memory transfer. And the learning behavior prediction accuracy is obviously improved. A sequence sampling strategy is adaptively adjusted by using a deformable sampling mechanism, feature representations of different time granularities are extracted in combination with a multi-scale convolution encoder, and a short-term memory transfer module is introduced to simulate a cognitive rule of human memory, so that complex behaviors in a learning process can be effectively captured. According to the method, multilevel modeling from microscopic local learning behaviors to macroscopic knowledge state evolution is realized, more accurate and reliable technical support is provided for a personalized learning system, the learning development trend of students can be scientifically predicted, and a powerful basis is provided for teachers to formulate accurate teaching intervention strategies.

A computer-implemented method is provided performed by a client computing device for decentralized learning based on local learning at the client computing device is provided. The method includes training a local M, model based on an activation function using a local parameter set and a reference parameter set to obtain a setting for respective 5 local parameters in the local parameter set that minimizes a training loss wherein the activation function preserves agreements and discourages disagreements between the local parameter set and the reference parameter set. The method further includes sending the trained local ML model to a server computing device. The method further includes receiving, from the server computing device, a global ML model that meets a convergence criterion. A 10 method performed by a server computing device, and related methods and apparatuses are also provided.

The application provides a lightweight remote sensing change detection method based on a non-back propagation learning, comprising the following steps: uniformly cutting remote sensing images before and after changes to be inferred to obtain a sub-image set; using a block dimension-based affinity matrix method, selecting changed sub-images before and after changes and unchanged sub-images before and after changes from the sub-image set as training samples of a neural network; using the training samples, fine-tuning the neural network by using a local learning method, assigning a loss function to a convolutional layer to be trained for fine-tuning, keeping parameters of other convolutional layers unchanged, and making the other convolutional layers perform forward propagation to obtain a fine-tuned neural network; inputting the remote sensing images before and after changes to be inferred into the neural network for inference, and binarizing an inference result to obtain a binary change result. The application adopting the above scheme is suitable for heterogeneous remote sensing image change detection when the computing resources are limited.

A federated learning method includes sampling at least one edge device to perform a federated learning based on a local data of each of edge devices, aggregating a result of a local learning by performing the local learning of the edge device which is sampled, generating a global weight based on the result of the local learning which is aggregated, and broadcasting to each of the edge devices.

Each local learning device included in a training system learns a subset of correction weighting factors for correcting part of a set of weighting factors of a basic model that is a base of a generation model, using a set of local training data, generates distribution data representing distribution of a feature of data included in the set of local training data, and transmits the learned subset and the distribution data to a server. The server generates a set of artificial training data, based on the distribution data received from each local learning device; the set of artificial training data reproduces distribution of a feature represented in the distribution data. With the set of artificial training data, the server trains a gate network for selecting a subset of correction weighting factors to be used.

The invention relates to the technical field of intelligent wearable equipment, and particularly discloses an electric heating bra multi-point temperature control system based on intelligent temperature balance control. The system comprises a flexible substrate and heating array, a multi-mode sensing and data acquisition unit, a local edge calculation and control decision unit and a power management and wireless communication unit. The local edge calculation and control decision-making unit operates locally at a user terminal, dynamically adjusts the power of each heating sheet through a double-layer balance control strategy based on sensing data, and optimizes control parameters in a personalized manner by using a local learning module integrated with differential privacy. The system realizes multi-point temperature adaptive equalization control and personalized comfort improvement on the premise of strictly protecting user privacy.

The application relates to the technical field of microprocessors, and specifically provides a local learning and adaptive inference method and system. First, millimeter wave radar data collected is used for data cleaning, then a TinyML model is trained, and finally the model is deployed to a microprocessor. Meanwhile, millimeter wave radar data is transmitted to a fully connected neural network layer in the microcontroller as input features, then each input feature data corresponds to a current microprocessor device state parameter as a label of the input feature data, with the continuous increase of the input feature data and the corresponding label, the weight parameters of the fully connected neural network layer are optimized according to the actual state of the device, and finally the updated parameters are updated to the TinyML model for local learning and adaptive inference. Compared with the prior art, the application reduces the dependence on high-performance edge devices, and reduces the security problems in the data transmission process of the traditional scheme.

This system provides a learning system that can appropriately train the entire generative model without removing the set of training data used to train a part of the generative model from the training device. [Solution] Each local learning device 2 of the learning system 1 learns a subset of correction weight coefficients that corrects a part of the set of weight coefficients of the basic model that forms the basis of the generative model using a set of local training data, generates distribution data that represents the distribution of features of the data included in the set of local training data, and transmits the learned subset and distribution data to the server 3. Based on the distribution data received from each local learning device 2, the server 3 generates a set of pseudo-training data that reproduces the distribution of features represented in that distribution data, and learns a gate network for selecting the subset of correction weight coefficients to use using the set of pseudo-training data.

Disclosed herein is performing model parallel training with local learning in a processor-based device. In some aspects, a distributed training system includes a plurality of processor-based nodes, where each node is configured to receive input data and process the input data using a plurality of modules including different subsets of layers of a large language model (LLM). Each node computes a local loss function and computes a local gradient based on the local loss function without communicating the local gradient to other nodes of the plurality of processor-based nodes. Each node is also configured to forward local node data to a next subsequent node, where the local node data includes one or more of an intermediate representation and a logical value. Each next subsequent node includes a fusion block configured to fuse the local node data of a previous node with the local node data of the next subsequent node.

The application relates to the technical field of intelligent application of communication networks, in particular to a spatial channel dynamic characteristic prediction system based on deep learning. The system comprises the following modules: a non-stationary topology detection module, which is used for obtaining broken-link communication data; non-stationary topology detection is carried out according to the broken-link communication data, and non-stationary topology data is obtained; a subset cluster feasible communication domain construction module, which is used for carrying out subset cluster feasible communication domain construction according to the non-stationary topology data, and obtaining subset cluster feasible communication domain data; a communication uncertainty prediction module, which is used for carrying out communication uncertainty prediction according to the subset cluster feasible communication domain data, and obtaining communication uncertainty data; a cross-domain opportunity action learning module, which is used for carrying out broken-link topology local learning according to the communication uncertainty data, and obtaining local prediction data; cross-domain opportunity action learning is carried out according to the local prediction data, and broken-link cross-domain reconnection data is obtained. The application can realize dynamic reconnection strategy optimization under the environment of broken-link communication of a UAV cluster.