385 results about "Auto encoders" patented technology

A system and method for multi-scale temporal attention processing in multimodal technology deep learning systems. This system processes time-series, textual, sentiment, and structured tabular data across three hierarchically-organized temporal streams—quarterly, weekly, and intraday levels—with bidirectional cross-temporal information flow. Scale-specific attention mechanisms are optimized for respective temporal granularities, while an adaptive controller dynamically weights each temporal level based on real-time market volatility indicators. A multi-scale fusion processor integrates attention-weighted representations to generate temporally unified representations preserving both short-term market dynamics and long-term trends. This approach enables superior forecasting and risk assessment by leveraging temporal correlations across multiple time scales while automatically adapting to changing market conditions. The system facilitates interpretable AI analysis through attention visualization and enables synthetic scenario generation for model testing.

A method is provided for enhanced data routing in neural networks using Self-Organizing Maps (SOM) integrated with Autoencoder-GAN. The method comprises training an autoencoder to encode input data into a latent space representation; applying a Self-Organizing Map (SOM) to organize the latent space representation into a topological map; refining the latent space representation using a Generative Adversarial Network (GAN), wherein the generator generates enhanced latent space representations and the discriminator evaluates their quality; using the refined latent space representations to update the SOM topology dynamically; generating routing coefficients based on the updated SOM topology to guide data routing in a capsule network; and dynamically adjusting routing within the capsule network using the generated routing coefficients to enhance performance based on the refined latent representations.

A system and method for adaptive real-time multi-modal compression with dynamic resource allocation provides intelligent compression optimization based on continuously monitored device conditions. The system monitors battery level, CPU utilization, and memory availability while classifying incoming multi-modal data streams comprising image, audio, text, and sensor data to determine processing priorities. Multi-objective optimization balances compression efficiency, reconstruction quality, and energy consumption using evolutionary algorithms that generate optimal parameters for an adaptive variational autoencoder. The autoencoder features dynamically selectable processing complexity, adjustable latent space dimensionality, and modality-specific processing layers. The system automatically switches between operational modes including emergency mode triggered by resource constraints, which applies maximum compression settings and intelligent data triage. Continuous learning adapts compression parameters based on observed performance outcomes, improving future optimization decisions. The system enables homomorphic operations on compressed data and provides enhanced compression performance under varying resource constraints across diverse edge computing applications.

A neural network system is provided. The system includes an autoencoder configured to encode input data into a latent space representation; a generator neural network configured to receive a noise vector and the latent space representation and output a set of routing coefficients; a discriminator neural network configured to evaluate the effectiveness of the routing coefficients by measuring the performance of a capsule network utilizing said routing coefficients; and a capsule network comprising a first capsule layer and a second capsule layer, wherein the routing coefficients are used to dynamically route outputs from the first capsule layer to the second capsule layer.

The invention relates to the technical field of fault detection, and particularly discloses a prediction reconstruction framework causal perception space-time network for explaining anomaly monitoring in a complex industrial process, comprising the following steps: S01, constructing graph data E (V) and a causal graph; automatically adjusting the fusion proportion of the time-frequency characteristics according to the data characteristics so as to ensure that the model can comprehensively capture the information of the data in the time domain and the frequency domain; secondly, introducing a residual image attention network (RGAT), and converting the image data E (V) into image structure data G (S (V), E (V)); and S03, reconstructing a prediction error by adopting a variational automatic encoder (VAE), learning an error mode of normal data, providing an anomaly judgment AD (V) for anomaly detection, analyzing a causal relationship between data in combination with a causal graph, and positioning an anomaly reason according to an anomaly score, so as to form a prediction result. The network solves the problem that a traditional monitoring network is high in false alarm rate.

The invention belongs to the technical field of artificial intelligence and information security crossing, and discloses a multi-modal false news detection method based on decoupling and cross-modal clue mining, and the method comprises the steps: employing a pre-training comparison model to extract text and image features in news, and mapping the text and image features to a unified embedding space; designing a multi-modal decoupling automatic encoder, decomposing each modal feature into modal common and modal specific representation, and obtaining a pure decoupling feature through confrontation constraint and reconstruction loss; a cross-modal clue mining module is constructed, modeling complementation is carried out from non-unhooked features, and conflicting semantic clues are enhanced; a self-adaptive attention aggregation module is introduced, and the modal features and the semantic clues are dynamically weighted and fused; and inputting the aggregated representations into a classifier for false news identification. According to the method, through modal decoupling and multi-source clue modeling, modal redundancy and specific information are effectively separated, association and conflicts between images and texts are deeply mined, and the detection accuracy is remarkably improved.

Provided is a method for token-based representation and learning of robotic perception data based on a graph neural network, comprising: obtaining a plurality of types of perception data of a robot; performing token-based representation according to types of the plurality of types of perception data; constructing an initial feature graph based on the plurality of types of perception data after the token-based representation; learning a compact representation of the initial feature graph based on an autoencoder and reconstructing a graph structure; after the autoencoder completes learning of the graph structure, fixing the graph structure; and converting the plurality of types of perception data into node feature vectors, constructing a feature graph based on the graph structure, and performing numerical encoding on each of the node feature vectors by utilizing the graph neural network to obtain a representation of high-dimensional feature vectors of the plurality of types of perception data.

The invention discloses an auxiliary film reading method and system based on artificial intelligence, and the method comprises the steps: 1, collecting a pathological WSI, an electronic medical record, detection data and equipment parameters, correcting the equipment difference through adaptive dyeing normalization, and constructing a structured data package associated with an ID-timestamp of a patient; 2, developing a dynamic branch CNN, migrating teacher model knowledge through knowledge distillation, and introducing federated learning; step 3, the edge generates a thermodynamic diagram to mark a suspicious area, and the cloud outputs a structured report; step 4, constructing a normal tissue feature space by the variational auto-encoder, detecting abnormal slices and triggering expert re-checking; a reverse automatic encoder generates a pseudo-health image to compare and position a pathological area, and dynamic weight adjustment balances the federal learning convergence speed; 5, integrating the thermodynamic diagram, the gene data and the clinical indexes by a three-dimensional platform, and supporting multi-dimensional superposition display; webGL realizes browser end rendering, and NLP automatically generates a report abstract marked with a key evidence chain and is in butt joint with an international diagnosis and treatment guide.

The invention relates to the technical field of air target prediction, in particular to an air target trend prediction method, which comprises the following steps: S1, multi-source heterogeneous data adaptive fusion filtering processing; s2, manifold learning is constructed in the high-dimensional spatial-temporal feature space; s3, performing semantic modeling on the dynamic behavior pattern recognition intention; and S4, multi-dimensional threat situation assessment warfare area modeling is carried out. According to the method, high-precision space-time synchronization and noise suppression of radar sensor data, infrared sensor data and other sensor data are achieved through the multi-source heterogeneous data self-adaptive fusion filtering technology, target micro-Doppler features are effectively reserved, missing data are repaired, and through the combination of third-order Savitzky-Golay differential filtering and short-time Fourier transform, high-precision space-time synchronization and noise suppression of radar sensor data and infrared sensor data are achieved. An 18-dimensional compression feature space containing kinematics and electromagnetic characteristics is constructed, a nonlinear topological relation is reserved through t-SNE and an automatic encoder, the signal-to-noise ratio and feature expression capacity of original data are remarkably improved through the function, and a high-precision and low-redundancy input basis is provided for follow-up behavior recognition and prediction.

Techniques for latent space based steganographic image generation are described. A processing device, for instance, receives a digital image and a secret that includes a bit string. A pretrained encoder of an autoencoder generates an embedding of the digital image that includes latent code. A secret encoder is trained and utilized to generate an embedding of the secret to act as a latent offset to the latent code. The processing device leverages a pretrained decoder of the autoencoder to generate a steganographic image based on the embedding of the secret and the embedding of the digital image. The steganographic image includes the secret and is visually indiscernible from the digital image. Further, the processing device is configured to recover the secret from the steganographic image, such as by training and leveraging a secret decoder to extract the secret.

The invention discloses a few-sample target detection method and system based on an aggregation variational prototype. The method comprises the steps of constructing a data set containing a base class and a new class, dividing the data set into a support set and a query set, generating a class prototype by utilizing a P-VAE module in combination with CLIP semantic features and a feature discriminator, realizing bidirectional fusion of the support set and the query set features by means of an MFM module, fusing the query features and the class prototype, and inputting the fused query features and the class prototype into a detection head to complete target detection. The system comprises a data set construction module, a priori variational automatic encoder P-VAE module, a mutual fusion module MFM, a feature fusion module and a target detection module. According to the scheme, by introducing semantic priori, optimizing prototype generation and feature interaction, the problems of data imbalance and insufficient new class feature representation in a few-sample scene are solved, improvement of new class detection precision is verified on PASCAL VOC, MS COCO and other data sets, and an effective solution is provided for target detection of sample scarce scenes such as medical images and rare species monitoring.

Systems and methods for detecting and diagnosing machine faults are discussed. An exemplary system includes at least one sensor node to sense a signal indicative of an operation status of a machine part, and a machine health analyzer circuit to generate a computational machine fault model comprising an autoencoder (AE) network and an associative module. The AE network encodes the sensed signal into signal features in a latent feature space, and decodes the signal features to produce a reconstructed signal. The associative module transforms the encoded signal features into an associative output using a dynamically updatable codebook. The machine health analyzer circuit detects a presence or absence of fault in the machine part based on reconstruction losses determined respectively from the reconstructed signal and the associative output. The detected fault can be presented to a user or to a process such as fault diagnosis or fault correction.

In various examples, self-supervised learning may be used to pre-train an encoder network of a masked prediction model to reconstruct masked regions of an input representation of 3D detections such as LiDAR point cloud(s). Spatial and / or temporal masking may be applied to a projected representation of 3D detections (e.g., a two-dimensional (2D) projection image), and the masked prediction model (e.g., a masked auto-encoder or joint-embedding predictive architecture) may be used to reconstruct a representation of the masked regions (e.g., reflection characteristic(s) stored in corresponding pixels or cells of the projected representation, a latent representation of the reflection characteristic(s)) during iterations of self-supervised learning. As such, the pre-trained encoder network of the masked prediction model may be used as a foundation model and fine-tuned with a task-specific output head or its pre-trained weights may be used to initialize a task-specific model.

A system is provided for dynamic analysis and verification of smart contracts. The system includes an autoencoder configured to preprocess smart contract code to reduce noise and highlight critical features; a capsule network configured to analyze the preprocessed smart contract code, capturing hierarchical relationships and dependencies within the code; a generative adversarial network (GAN) configured to generate optimal routing coefficients for the capsule network, enhancing the efficiency and accuracy of the analysis; and a blockchain-based platform for deploying and executing smart contracts, wherein the platform utilizes the capsule network to continuously monitor the smart contracts for anomalies during execution.

A generative machine learning model, such as a convolutional neural network (CNN), can be trained with solutions from a topology optimization solver for a solution for a topology of a set of structures so that the generative machine learning model can generate a plurality of alternative designs for a structure that are alternative topology optimizations (for the structure) for a set of initial setup parameters. The generative model when being trained includes a generative network and a discriminator network. The generative model can be trained using outputs from a CNN autoencoder for densities and a CNN autoencoder for strain energies.

The invention provides a construction and use method of a potential diffusion model for SAR image super-resolution. The construction and use method comprises the steps of obtaining an original SAR image and inputting the original SAR image into a real degradation model to generate a degraded SAR image; inputting the degraded SAR image into an automatic encoder to generate a structure-enhanced submerged space feature map; and inputting the structure-enhanced latent space feature map and the degraded SAR image into a potential diffusion model to generate an SAR super-resolution image. The method has the beneficial effects that a two-stage training strategy is adopted, different optimization targets are focused in stages, the training efficiency is improved, and meanwhile, the learning ability of the model to SAR image features is enhanced; sAR imaging key degradation factors are comprehensively covered, so that a generated low-resolution sample is closer to a real scene, high-quality data support is provided for model training, and model learning is prevented from being separated from an actual degradation rule; sAR specific interference such as speckle noise is effectively simulated, and the anti-noise training effect of the model is enhanced.

The invention discloses a lime kiln monitoring system based on multi-source data dynamic fusion, which is characterized by comprising a thermal imager, an infrared sensor, a pressure sensor, a gas analyzer and an edge computing node, the edge computing node is configured with a Raspberry Pi platform and an LSTM neural network accelerator and is used for executing multi-source data preprocessing and feature compression, and the Raspberry Pi platform and the LSTM neural network accelerator are connected with the thermal imager. The system also comprises a working condition analysis system, and the working condition analysis system comprises a ring formation prediction model based on ST-GCN, a combustion optimization model based on DDPG, and an anomaly detection model based on an isolated forest and automatic encoder hybrid model. The invention aims to solve the limitation of a traditional monitoring system, realize real-time monitoring and accurate control, improve the equipment health management and predictive maintenance capability, optimize waste heat recovery and energy utilization, and enhance man-machine interaction and operation convenience.

A method of training a machine learning model for reconstructing video or audio data based on neuroimaging data of a subject is provided. The method includes: training a neuroimaging data encoder based on neuroimaging data from a neuroimaging training dataset for generating neuroimaging data embeddings; and training a diffusion model based on the neuroimaging data embeddings generated by the neuroimaging data encoder as conditions on the diffusion model. In this regard, the diffusion model is trained to reconstruct video or audio data based on neuroimaging data embeddings of neuroimaging data of a subject obtained in response to a visual or audio stimulus. The neuroimaging data encoder includes a masked autoencoder. The above-mentioned training the neuroimaging data encoder includes training an encoder of the masked autoencoder based on the neuroimaging training dataset using unsupervised learning with masked data modeling for generating the neuroimaging data embeddings. The unsupcrviscd learning with masked data modeling includes generating neuroimaging data embeddings from neuroimaging data from the neuroimaging training dataset, masking a portion of the neuroimaging data embeddings into masked neuroimaging data embeddings and training the masked autoencoder to recover the masked neuroimaging data embeddings. There is also provided a method of using the machine learning model trained for reconstructing video or audio data based on neuroimaging data of a subject.

The invention discloses a converter endpoint carbon temperature dynamic adaptive prediction method based on meta-learning SAE, and the method comprises the steps: firstly carrying out the clustering of historical steelmaking data through employing a Wasserstein distance weighted Dirichlet process Gaussian mixture model (WDPGMM), and automatically dividing a plurality of working condition modes; then, constructing a MetaSAE model for the data of each mode, and training the MetaSAE model to improve the generalization ability of the model to different working conditions; and finally, according to the posterior probability of a new sample of a to-be-detected heat, judging the working condition mode to which the new sample belongs, selecting the most similar sample subset from the historical data of the corresponding mode by using the mutual information weighted JS divergence, and performing real-time fine adjustment on the MetaSAE model, thereby realizing the dynamic prediction of the molten steel end point carbon content and temperature of the new sample. The method can adapt to data distribution changes caused by multiple working conditions and sensor drifting in the steelmaking process, and continuous real-time accurate prediction of the carbon content and temperature of the molten steel is achieved.

This invention provides a multi-UAV path planning and power allocation method based on an autoencoder, relating to the field of UAV technology. This invention studies the power allocation and path planning problem in multi-UAV scenarios, aiming to maximize the average data transmission rate of user devices in a stochastic user environment. To this end, this invention proposes a deep reinforcement learning framework for distributed multi-agent environments, providing a solution for autonomous UAV collaboration. This framework uses a self-supervised representation learning task based on an autoencoder to learn the common communication basis of UAVs. On this basis, UAVs can understand and convey information observed by each other, improving their information acquisition capabilities under constrained observation conditions. This method is applicable to fully distributed architectures and does not require additional auxiliary information, achieving efficient information transmission and communication between UAVs, thereby effectively improving the service efficiency of UAV networks.

Generally discussed herein are devices, systems, and methods for. A method can include obtaining a normalizing autoencoder, the normalizing autoencoder trained based on first data samples of a template person and second data samples of a variety of people, normalizing, by the normalizing autoencoder, an input data sample by combining dynamic characteristics of a person in the input data sample with static characteristics in the first data samples, to generate normalized data, and providing the normalized data as input to a classifier model to classify the input data based on the dynamic characteristics of the input data and the static characteristics of the first data samples.

The invention discloses a GRU-based vehicle-mounted CAN intrusion detection method and system, a storage medium and a computer system. According to the method, an automatic encoder (AE) is introduced to deepen the understanding of a model on input sequence characteristics, a sliding window is used for selecting batch CAN data to be preprocessed to obtain 13-dimensional time sequence data, and a scalar value within the range of [0, 1] is obtained through processing of the encoder, a GRU, a decoder, a full connection layer and a sigmoid activation function and used for classification of abnormal data. The Conv1D is used as a hidden layer, and compared with two-dimensional convolution, the one-dimensional convolution parameter quantity is smaller, and the calculation is simpler and more convenient. An attack message and a normal message can be completely distinguished, the precision and the accuracy rate reach 100%, and the precision in Fuzz detection is 0.9983; compared with the prior art, the method has high accuracy and reliability in the aspect of intrusion behavior detection, can effectively identify most intrusion events, and can keep a relatively low overall error rate, so that good balance between safety and availability is realized.

Various systems and methods are presented regarding detecting data drift. The data of interest can be batches of utterances received at an interface (e.g., a chatbot). The batches of utterances can be compared with topics present in training data utilized to train a data classifier (e.g., an autoencoder), wherein topics identified in the batches of utterances that are not present in the training data can be considered to be novel topics. The greater the presence of novel topics in a batch of utterances, the greater the divergence of the batch of utterances from the content of the training data. The novel topics can be identified and subsequently applied to the training data such that the data classifier can be re-trained with the novel topics, thereby causing the data classifier to be contemporaneous with the novel topics. In an embodiment, the utterances can be short streams of text, symbols, and suchlike.

A multivariate flight data anomaly detection and location method based on a reconstruction model comprises the following steps: a data preprocessing step: collecting unmanned aerial vehicle (UAV) flight data, dividing the original input data into a training set and a test set according to a preset ratio, and normalizing the data; a model building step: building a neural network reconstruction model based on DCANNs; an autoencoder module compressing the input data into latent variables through an encoder, and then decoding z back into the original data space by a decoder to obtain reconstructed data; an adaptive anomaly detection step: calculating an anomaly threshold using the reconstructed residuals of the training set, and achieving anomaly detection by comparing the anomaly score of the test set with the threshold; and an anomaly location step: locating anomaly parameters by calculating and analyzing the statistical features of the residuals. The present invention aims to solve the problems of insufficient feature extraction in traditional deep learning models, the inability of fixed statistical thresholds to adapt to dynamic changes in flight data, and the inability to locate the specific source of anomaly parameters.

The invention discloses a silicon wafer edge defect detection method and device under a special chamfering process, and relates to the technical field of semiconductor manufacturing. The method comprises the following steps: obtaining a defect-free silicon wafer edge image under a special chamfering process, and obtaining an edge feature tensor through space unification, channel splicing and residual compression by taking the multi-layer and multi-scale features of the defect-free silicon wafer edge image in pre-training WideResNet50 as supervision; training and freezing an automatic encoder by taking the tensor as a target, and training a convolutional neural network by taking the output of the encoder as supervision; then, respectively carrying out edge defect detection on the to-be-detected image by utilizing an automatic encoder with frozen parameters and a trained convolutional neural network to obtain two groups of defect feature maps; generating defect feature maps by calculating pixel-by-pixel differences of the two groups of feature maps in channel, width and height dimensions, and mapping the defect feature maps into defect scores; and highlighting and positioning a defect area according to the defect score, and outputting a silicon wafer edge defect detection result. The method effectively solves the problem of insufficient industrial field defect labels.

The invention relates to an abnormal traffic detection method and device, electronic equipment and a storage medium, and relates to the technical field of network security, and the method comprises the steps: carrying out the preprocessing of original network traffic data, and obtaining a high-dimensional feature vector; inputting the high-dimensional feature vector into a stack sparse automatic encoder for feature dimension reduction processing to obtain a low-dimensional feature vector; inputting the low-dimensional feature vector into a recurrent neural network for forward and backward parallelization processing to obtain a prediction probability value; if the prediction probability value is larger than or equal to the preset probability threshold value, it is judged that the original network flow data is abnormal flow, by adopting the technical scheme, manual feature design is not needed, manual intervention is reduced, changes of the network environment and attack means can be better adapted, frequent feature engineering process adjustment in a traditional method is not needed, and the method is suitable for large-scale popularization and application. The complexity and the cost are reduced, the error of feature extraction is reduced, and the detection accuracy is improved.

The present invention is applicable to the field of EEG emotion recognition technology, and provides an EEG emotion recognition method based on deep learning, including: first preprocessing the EEG signal, including removing baseline noise, standardization, bandpass filtering and signal segmentation; then establishing a deep learning model, the deep learning model sequentially including an autoencoder, a dynamic graph convolutional neural network, a Transformer and a classifier; then using the cross-entropy loss function combined with regularization to train the deep learning model; finally, based on the trained model, performing emotion recognition on the preprocessed EEG signal and outputting the result. The present invention realizes the collaborative extraction of dynamic spatial features and long-range time-dependent features in EEG signals by fusing multiple deep learning models. The present invention greatly improves the accuracy of EEG emotion recognition, and provides more reliable and efficient emotion recognition technology support for human-computer interaction fields such as depression assessment and affective disorder treatment.

The embodiment of the invention provides an air cooling island anti-freezing control method and system based on AI prediction and dual-stage optimization, and the method is characterized in that the method comprises the steps: carrying out the analysis of collected air cooling island data through an AE model, evaluating the freezing probability of each fan region in a given time window, and calculating the freezing probability of each fan region; according to the freezing probability, the freezing risk of each fan area is divided into three levels of low risk, medium risk and high risk; when the freezing risk is identified as a medium risk or a high risk, establishing a first optimization model of fan speed change and temperature change of a corresponding temperature measurement zone, and performing first-stage optimization by taking air cooling island freezing risk minimization as a target; when the first optimization model cannot meet the anti-freezing requirement, establishing a second optimization model of the rotating speed of the reverse fan, the start and stop of the vacuum pump and the temperature of the temperature measuring zone, and performing second-stage optimization; the optimization result is converted into a control instruction capable of being issued, and feedback signals are recycled to correct the automatic encoder model in a closed loop mode.

The provided system and methods describe a Multi-factor Authentication (MFA) kiosk that utilizes various sensors to capture biometric, behavioral, and physiological data for authentication. The kiosk includes a user interface, a set of sensors, and services such as kiosk management, rules configuration, sensor management, and an authentication service. The sensors, both integrated and external, gather diverse data, including facial recognition, fingerprint scans, voice recognition, gait analysis, and more, constructing a physical profile for authentication. The system incorporates a rules service for configuring authentication policies and a sensor management service to optimize sensor performance. Authentication service uses a scoring model, potentially a deep learning algorithm like an autoencoder, to generate an authentication score based on inputs from sensors, rules, and previous attempts. Security measures include encryption, isolation of components, and compliance with data protection regulations. A plurality of MFA kiosks may form an authentication network.

The invention relates to an aero-engine fault diagnosis method and device based on tensor decomposition and a medium, and the method comprises the steps: collecting the gas path parameter time sequence data of a plurality of parts of an aero-engine, and carrying out the time alignment and working condition label labeling; for the multi-parameter data of each component, a kernel principal component analysis and automatic encoder fused feature extraction method is adopted, and parameters of different components are unified into fixed dimension features; the feature matrixes obtained after feature extraction of all the components are combined, and a time-component-feature third-order tensor is constructed; training a high-order singular value decomposition model based on the engine health monitoring data; fault diagnosis is carried out by calculating the reconstruction error of the test data and the monitoring data; and fault positioning is realized by combining core tensor difference analysis. According to the method, the problem of feature fusion of engine multi-source heterogeneous data under variable working conditions is solved, and the engine fault detection sensitivity and positioning precision are improved.