251 results about "Learning network" patented technology

A system for compiling a machine learning network for execution on a plurality of interconnected processing elements performs the following. It selects splits (Ps,Qs,Ks) for individual layers of the machine learning network based on (a) costs of data transfer between processing elements assuming that inputs and outputs of individual layers have a same spatial split (Ps,Qs), and (b) costs of data transfer between processing elements resulting from differences in the spatial splits for inputs and outputs of individual layers. In addition, it selects the same splits (Ps,Qs,Ks) for layers that have the same spatial size P×Q. It then generates a computer program that allocates computations for executing the machine learning network to the processing elements according to the selected splits for the individual layers.

A method for privacy preservation for machine learning networks includes splitting a trained neural network into a first part and a second part. The first part is a privacy preservation (PP) encoder and the second part is a PP machine learning (ML) model. The method further includes retraining the PP encoder and the PP ML model. The method can improve artificial intelligence (AI) systems, optimize performance and support decision making in use cases including, but not limited to medical / healthcare, fraud detection, image recognition, predictive maintenance and connected vehicles.

An image processing apparatus applies an image to a first learning network model to optimize the edges of the image, applies the image to a second learning network model to optimize the texture of the image, and applies a first weight to the first image and a second weight to the second image based on information on the edge areas and the texture areas of the image to acquire an output image.

This invention discloses a method for optimizing the topology of a 3D on-chip network (SoC) for high-speed data acquisition systems. Addressing the problem that general-purpose 3D SoC architectures are difficult to adapt to the communication characteristics of high-speed data acquisition systems, leading to high transmission latency and redundant link resources, this method models the network topology adjustment process as a Markov decision process and employs deep reinforcement learning to achieve adaptive optimization of the topology. The method uses a general-purpose 3D SoC as the initial architecture, constructing a state representation that includes node connection features, average network latency, longest path latency, and link area; defining a pruning and regrowth action space; designing a reward function that integrates changes in average latency, longest path latency, and link area; and using a deep Q-learning network with a multilayer perceptron structure for policy training. To improve learning efficiency, an action candidate set is constructed and combined with a two-layer greedy policy to achieve fast and effective search. Through iterative learning and topology updates, a 3D SoC topology that better suits the communication load of high-speed data acquisition systems can be obtained, reducing average transmission latency and link area overhead while ensuring connectivity.

This invention relates to the field of ischemic injury model construction technology, and particularly to a method for constructing a mouse ischemic injury model for evaluating the protective effect of GDF-15. The method includes acquiring a dataset of mouse physiological parameters to determine physiological feature vectors, thereby determining an individual vulnerability index and performing similarity matching with the vulnerability index of a specific individual. Based on the matching results, a matching sample set is constructed, and then an ischemic injury model is constructed using a deep learning network model. The model is then dynamically simulated using the specific individual vulnerability index, specific intermediate monitoring data, and specific infarct area percentage data to output a simulated protective effect index to determine if the model is qualified. If unqualified, the similarity matching threshold is adjusted based on the degree of deviation, and the matching and model construction are repeated. This invention significantly improves the uniformity and construction success rate of the ischemic injury model through individualized matching and dynamic feedback optimization, providing a reliable model for evaluating the protective effect of GDF-15.

The application discloses an urban waterlogging intelligent prediction method and system based on a physically constrained enhanced video generative adversarial network. The method comprises the following steps: constructing a meteorological geographic data set containing static terrain data and dynamic rainfall data; generating waterlogging simulation data based on a physical numerical simulation model, and matching the waterlogging simulation data with the meteorological geographic data set into a comprehensive data set; constructing a physically constrained enhanced video generative adversarial network model, wherein the model comprises a generator and a discriminator; constructing a multi-angle loss function system that fuses an adversarial loss, a time sequence consistency loss, a data supervision loss and a water quantity balance constraint physical loss; and alternately optimizing the generator and the discriminator until the model converges, and outputting a spatiotemporally continuous waterlogging inundation depth prediction result. The application explicitly embeds a physical hydrological law into a deep learning network, and eliminates the physical logic contradiction of a pure data-driven model in continuous spatiotemporal sequence prediction through water quantity balance constraint.

Automatically generating a structured medical note during a remote medical consultation using machine learning. A provider tele-presence device may receive audio from a medical provider. A medical documentation server may be coupled to the network. A machine learning network receives audio data from the provider tele-presence device, the machine learning network generating a structured medical note based on the received audio data, and wherein the structured medical note is stored in the medical documentation server in association with an identity of a patient.

The application relates to a sentiment analysis method based on multi-source knowledge and multi-granularity image-text features, which comprises the following steps: step A: collecting text and picture data, identifying and marking aspect words and their sentiment polarities in the text to form a training set; step B: constructing a deep learning network model based on external multi-source knowledge and multi-granularity image-text features, obtaining knowledge-enhanced text feature representation and image feature representation by using AMR graphs and image labels, and obtaining multi-granularity text-visual fusion feature representation by using the syntactic dependency relationship of the text, the component tree structure of the text and the fine-grained image-text relationship matrix in combination with a multi-layer graph attention network; training the deep learning network model by using the training set; and step C: inputting the text data and the picture data into the trained deep learning network model in sequence, extracting aspect words in the text data, and predicting the sentiment polarities corresponding to the aspect words. The method and system are beneficial to improving the accuracy of sentiment polarity analysis.

The embodiment of the application provides a kind of based on deep reinforcement learning's mechanical arm autonomous dynamic obstacle avoidance method and system, belong to automation and robot technical field.The mechanical arm autonomous dynamic obstacle avoidance method includes: initialize current environment parameter and experience replay pool;Select the action currently required to be executed;The selected action is sent to server side, to complete the update of mechanical arm state;According to the updated mechanical arm state, update reward function value;The state before executing action, the action selected, the state after executing action and reward value are stored in experience replay pool;Whether the data amount in experience replay pool reaches minimum sampling number is judged;Experience replay pool is used to train deep learning network, and the obstacle avoidance accuracy of deep learning network is determined;Whether the difference between the obstacle avoidance accuracy in this round iteration and the obstacle avoidance accuracy in last round iteration is greater than preset value is judged;In the case where it is judged less than preset value, output deep learning network.

This invention provides a method for early warning and self-healing of material blockages on production lines that integrates the Internet of Things (IoT). It relates to the fields of industrial intelligent manufacturing and automation technology. The method involves deploying multiple types of sensors at key equipment nodes on the production line to collect multi-source heterogeneous data. This data is then transmitted to edge computing nodes for preprocessing, outputting a multimodal feature dataset. A production line blockage early warning model is constructed based on this multimodal fusion deep learning network model. Blockage monitoring is performed using this model. Based on the monitoring results, corresponding self-healing strategies are executed, generating control commands that are sent to the production line actuators. The entire chain of data for each early warning and self-healing event, along with its final result labels, is collected to form a feedback dataset. This dataset is then used for periodic incremental training and optimization of the production line blockage early warning model. This invention achieves early warning and autonomous handling of blockages, improving production continuity and intelligence.

The present application provides a multi-modal deep learning network-based overlay mark asymmetry compensation method and apparatus. Manufacturing data of different modalities is acquired and then inputted into a multi-modal deep learning network for processing, so as to obtain a prediction error of overlay mark asymmetry, wherein the multi-modal deep learning network performs feature extraction on the manufacturing data of different modalities, performs feature fusion on vector features corresponding to different modalities, and associates and combines physical parameters of actual production in a manufacturing process with parameters observed by a device, so as to obtain the prediction error of overlay mark asymmetry, thereby compensating for an actual overlay mark in actual production. By introducing a multi-modal learning network, a variety of data features are fully mined and utilized, the generalization ability of a model is enhanced, and the intelligence of the device is realized by means of the machine learning technology, thereby providing an effective traceability and localization solution for process problems in wafer production.

A machine learning accelerator (MLA) implemented on a semiconductor die includes a computing mesh of interconnected compute elements. The compute elements execute a program of instructions to implement a machine learning network according to a static schedule for execution of the instructions. The compute elements include gap counters. The number of cycles between any two instructions (i.e., the gap count) in a statically scheduled program is known and fixed. A gap counter counts cycles during execution and must reach the expected gap count before the later instruction can be executed. Synchronization between different processing elements may be maintained by suspending counting for a period of time.

The application relates to the technical field of algorithm and network cooperative scheduling, and discloses an algorithm and power cooperative scheduling method and system based on multi-dimensional classification of task value and time delay sensitivity, which comprises the following steps: acquiring multi-source data streams of computing power tasks, power markets and network topologies; performing multi-dimensional classification processing based on a time delay sensitivity index and a comprehensive value score to generate a task type code; extracting power market environment features to dynamically generate a multi-target optimization weight vector, and screening network topology nodes to generate a candidate data center set; combining the code and the weight vector to calculate the comprehensive cooperative utility of each node, generate a to-be-verified scheduling decision, input the decision into a consortium chain network to perform a quoted price tolerance check and consensus determination, and generate a scheduling instruction after the check and determination; collecting actual operation parameters according to the instruction to construct an experience four-tuple, and inputting the experience four-tuple into a meta-learning network for parameter fine-tuning. Through multi-dimensional feature mapping and distributed checking, the application realizes closed-loop cooperative scheduling of heterogeneous computing power and dynamic power.

The application discloses an image denoising method and device, electronic equipment and storage medium, and relates to the technical field of artificial intelligence, wherein the image denoising method comprises the following steps: receiving a document image to be processed, and preprocessing the document image; removing first frequency noise in the preprocessed document image by using a preset deep learning network, extracting image features after noise removal processing to obtain a first image feature set, extracting image features conforming to second frequency noise in the document image by using a preset deep residual network to obtain a second image feature set, fusing the first image feature set and the second image feature set, inputting a feature map obtained after fusion into a target convolutional neural network, and outputting a target document image. The application solves the technical problem that image blurring is easily caused when a document image is denoised in the related art.

This invention relates to a multi-temporal radiation signal inversion method based on Conditional Transformer, belonging to the field of satellite remote sensing data processing / spatial information science and technology. The method includes: acquiring meteorological time-series data of the target area to establish a dataset; designing a Conditional Transformer deep learning network; combining the target time-series data and meteorological data encoded by time and meteorological code generators with the output of the Encoder layer, and inputting this combination into the Decoder layer to train the Conditional Transformer deep learning network to obtain a surface radiation signal inversion model; inverting surface radiation brightness temperature data and evaluating the inversion accuracy; efficiently capturing long-distance spatiotemporal dependencies of radiation signals; flexibly inverting radiation changes according to the target time and meteorological conditions; removing nonlinear interference from the atmosphere on radiation; and more accurately inverting the true radiation characteristics of the surface.

A cross-domain recognition and drift compensation method for electronic noses based on contrastive variational autoencoders (CVAEs) is proposed, belonging to the interdisciplinary field of electronic nose systems and artificial intelligence. The method involves constructing a labeled dataset, building an initial C-VAE network and establishing a basic loss function, connecting the initial C-VAE network with a classifier to form a joint learning network, introducing a classification loss function to construct a combined C-VAE loss function, and training to obtain an initial model. A target domain dataset is then constructed and input into the initial model to obtain the latent space vector of the target domain. Mean-variance transformation is used to achieve style transfer and data augmentation of source domain features. The latent feature structure of the target domain is optimized through various constraints. A joint training set is constructed, and the depth alignment of features between the source and target domains is achieved by constraining the distribution differences of features between the two domains. Training is performed under the joint loss function, and progressive scheduling is applied to obtain a cross-domain gas recognition model. This method effectively solves the nonlinear drift problem of electronic nose sensors.

This invention provides a method and system for fitting aspherical removal functions based on deep learning, relating to the field of optical technology. The method includes: obtaining at least two true removal functions for spherical workpieces with different radii of curvature, and multiple true removal functions for a continuously curvatured aspherical calibration workpiece at corresponding continuously changing curvature positions on its surface; discretizing each removal function in polar coordinates to obtain several data samples; constructing a deep learning network; obtaining a preset composite loss function; using several data samples, iteratively optimizing the network parameters of the deep learning network with the goal of minimizing the composite loss function, to obtain a trained deep learning network; inputting the polar radius, polar angle, and radius of curvature of any point D on the aspherical surface to be processed into the trained deep learning network to obtain the height distribution of the removal function at point D; this invention can significantly improve the processing accuracy and surface quality of aspherical optical elements.

This invention proposed a channel estimation scheme for wireless communication system in high mobility environment. Different from conventional channel estimation schemes only considered the consist geometry information channel, the proposed scheme is developed for a high-resolution system, which is more sensitive to the path change of the signal propagation. A Deep-Learning network first provide a rough prediction, then an improved Least Squares estimation method is adopted for find estimation.

This invention discloses a method, device, equipment, and medium for monitoring high physiological stress based on rPPG. First, a video sequence containing the user's face is acquired, and the region of interest (ROI) sequence is extracted and input into a deep learning network model. A spatiotemporal encoder extracts preliminary spatiotemporal physiological feature sequences from these sequences. Then, relying on an adaptive frequency-aware decoder, multi-scale parallel processing and dynamic feature fusion are performed on these preliminary spatiotemporal physiological feature sequences to output feature representations of abnormal high-frequency rhythmic events. Finally, a state determiner, combined with the abnormal high-frequency rhythmic event identification results, directly outputs a binary classification result indicating whether the user is in a state of high physiological stress. This invention effectively solves the problem of abnormal high-frequency rhythmic signal loss caused by existing fixed filters, achieving high-precision identification of abnormal high-frequency rhythmic events and accurately determining the user's state of high physiological stress.

This disclosure relates to the field of computer science, providing a method, apparatus, electronic device, and medium for training a multimodal contrastive learning model. The method includes: acquiring multimodal data including text data and image data; processing the multimodal data to obtain serialized marker data corresponding to the text data and image patch data corresponding to the image data; inputting the serialized marker data and image patch data into a feature encoder to obtain initial text features and initial image features; inputting the initial text features and initial image features into a preset routing-based expert layer to obtain multimodal features; inputting the multimodal features into an encoding layer to obtain encoded text features and encoded image features; and training a contrastive learning network based on the encoded text features and encoded image features to obtain a trained multimodal contrastive learning model. This disclosure allows a single model to handle multiple single / multimodal tasks simultaneously, avoiding catastrophic forgetting problems and improving work efficiency and training effectiveness.

The invention discloses a tar absorbent performance prediction and high-throughput screening method, system and equipment and a medium. The method comprises the following steps: constructing an absorption assistant structure library and a tar component structure library; respectively sampling the two structure libraries, establishing a water phase-absorption aid-oil phase model, and carrying out sufficient molecular dynamics simulation until a data set with a specified scale is formed; pre-training the deep learning network model; selecting an actual tar component, determining a tar component descriptor, and screening out an absorption assistant subset by using a pre-trained deep learning network model; the method comprises the following steps: determining absorption aids of a plurality of representative structures in an absorption aid subset, determining absorption rates through experiments, performing fine tuning training on a model to obtain a high-throughput screening model, and predicting all structures in an absorption aid structure library to obtain the absorption aid with the highest absorption rate. The method has the advantages of high-precision prediction, low experiment cost, high expandability and the like, and is suitable for rapid research, development and screening of the absorption additive material.

The application provides a light field multispectral temperature inversion system and method based on a physical information neural network, first constructs a deep learning network infrastructure, adopts a double-layer U-Net architecture to extract radiation and spatial distribution characteristics of data; then designs a physical information embedding module, modularizes a Planck radiation law, and guides the network to establish a physical correlation between radiation information and temperature; subsequently, light field multispectral radiation data are collected through experiments, the network is trained after data division and preprocessing are completed, and a mapping relationship between radiation characteristics and temperature characteristics is established; finally, the trained network is migrated to actual test data, and high-precision and rapid temperature inversion is realized. The application has both the advantages of deep learning in processing complex data and the constraint of a physical model, not only improves the precision and calculation efficiency of light field multispectral temperature inversion, but also enhances the applicability of the light field multispectral temperature inversion in a high-temperature complex environment.

This invention relates to the field of intelligent monitoring technology for power equipment. To address the problems of isolated cable condition monitoring parameters, difficulty in deeply correlating with fundamental insulation aging indicators, and fuzzy anomaly location, a multi-state low-power sensing system for distribution cables is disclosed. This system acquires cable sheath temperature, discharge waveform, and insulation loss data through an information acquisition module; a condition generation module processes the data to obtain comprehensive cable condition characteristics; a weight learning module uses a trained anomaly representation learning network to generate an anomaly weight spectrum reflecting the correlation strength between operating conditions and insulation loss; a source tracing and location module performs dual-source anomaly trajectory backtracking analysis on the original temperature and discharge data based on this weight spectrum, achieving precise location of the anomaly's physical location and identification of its inherent defect type; and a parameter optimization module generates and executes a control parameter adjustment scheme. This invention achieves deep correlation assessment of cable insulation status and precise physical-level diagnosis of anomaly root causes.

The application discloses a remote sensing image change detection method, system, device, medium and product based on a multi-task deep learning network model, relates to the field of remote sensing image processing, and comprises the following steps: acquiring double-time multi-source space data of a target area; the double-time multi-source space data comprises double-time optical images, double-time SAR images and a double-time digital elevation model; dividing the target area into multiple height levels according to the double-time digital elevation model and generating a height coding map; based on the double-time multi-source space data and the height coding map, a pre-trained multi-task deep learning network model is used to output a change detection map, a change type map and a height change map, so that remote sensing image change detection is realized; the multi-task deep learning network model comprises an encoder, an alignment and fusion module, a four-dimensional neighborhood difference convolution module and a stereo gradient enhancement decoder. The application can realize high-precision and quantifiable remote sensing image change detection in a stereo scene.

Computer systems and computer-implemented methods modify a machine learning network, such as a deep neural network, to introduce judgment to the network. A “combining” node is added to the network, to thereby generate a modified network, where activation of the combining node is based, at least in part, on output from a subject node of the network. The computer system then trains the modified network by, for each training data item in a set of training data, performing forward and back propagation computations through the modified network, where the backward propagation computation through the modified network comprises computing estimated partial derivatives of an error function of an objective for the network, except that the combining node selectively blocks back-propagation of estimated partial derivatives to the subject node, even though activation of the combining node is based on the activation of the subject node.

This invention belongs to the fields of medical image processing and deep learning technology. It discloses a pathological image classification method and system combining stable learning and hybrid enhancement. The method involves acquiring a pathological image dataset, dividing it into a training set, a validation set, a test set, and an external validation set, and preprocessing the dataset. A deep learning network combining stable learning and hybrid enhancement is constructed and trained using the training set. The optimal deep learning network model is obtained using the validation set. The test set and the external validation set are input into the optimal deep learning network model to output the pathological image classification results. This invention utilizes a well-fitting pathological image classification model to effectively improve the overfitting problem and weak recognition ability of traditional models for domain-biased data, enhancing the recognition accuracy of independent and identically distributed data, improving the robustness and generalization ability of the pathological image classification model, and increasing the diagnostic accuracy of pathological images.