10results about How to "Suppress overfitting" patented technology

The application provides a kind of method for enhancing robustness of pulse neural network based on neural activation and connection optimization, applied to artificial intelligence and neural network technical field, the method comprises: constructing pulse neural network model, the neuron of pulse neural network model adopts burst enhancement type pulse neuron, the input of pulse neural network model is image data, and the output of pulse neural network model is the image processing result corresponding to computer vision task;When the membrane potential of burst enhancement type pulse neuron exceeds the firing threshold, the part exceeding the membrane potential is converted into the pulse output within the burst window by quantization linear mapping function, and the number of pulse output is an integer between 0 and the preset maximum burst pulse number;When training pulse neural network model, add activation perception regularization term to total loss function;Through the application, the accuracy and robustness can be simultaneously improved while maintaining the high energy efficiency advantage of pulse neural network.

This invention discloses a video summarization method based on two-layer routed sparse attention and spatial pixel recalibration, belonging to the field of computer vision. The method includes: reading the input video and extracting frame-level feature vectors; constructing a video summarization generation model, performing channel enhancement on the frame-level feature vectors to obtain enhanced features; performing two-layer routed sparse attention and spatial pixel recalibration on the enhanced features, and fusing them to obtain fused features; inputting the fused features into an importance scoring regression network, outputting frame importance scores, and selecting frames to generate video summaries. This invention effectively fuses global dependencies and local details by combining two-layer routed sparse attention and spatial pixel recalibration, accurately identifying key segments in the video while reducing computational complexity. Experimental results based on the SumMe and TVSum benchmark datasets demonstrate that the method of this invention exhibits good performance in video summarization tasks.

The application provides a stratum profile prediction method and system based on limit indicators and integrated learning, which comprises the following steps: obtaining a set of discrete sampling points of a two-dimensional profile to be predicted, extracting spatial coordinates, natural water content, liquid limit and plastic limit; calculating Atterberg derivative indicators based on the liquid limit and the plastic limit, the Atterberg derivative indicators including plasticity index and liquidity index, and constructing a multi-dimensional feature vector in combination with the spatial coordinates and the soil indicators; constructing a multi-stage weak supervision labeling mechanism with the Atterberg limit indicators to generate weak supervision soil class labels; training an integrated learning classification model with the multi-dimensional feature vector as the input and the labels as the target training set; gridding the profile to be predicted, obtaining the physical property indicators of the grid nodes through spatial interpolation, and constructing grid node feature vectors; inputting the model to obtain predicted soil class labels, and outputting a two-dimensional stratum distribution matrix. The application integrates conventional soil limit indicators and spatial information, realizes automatic, stable and reasonable continuous two-dimensional stratum profile prediction, and has engineering rationality.

This invention discloses an EEG emotion recognition method based on multi-channel residual convolutional Transformer capsules, belonging to the field of artificial intelligence technology. The method acquires the user's EEG signal, including baseline and stimulus-evoked signals. After baseline correction of the stimulus-evoked signals, multi-scale spatiotemporal feature maps are extracted using a multi-channel residual convolutional network. These maps are then globally correlated and encoded using a Transformer network to obtain a global contextual feature sequence. Finally, these sequences are converted into primary capsule vectors and aggregated into emotion capsule vectors using a dynamic routing protocol. The emotion recognition result is output based on the magnitude of the emotion capsule vectors. By achieving multi-scale spatiotemporal feature fusion through multi-channel residual convolution, capturing global contextual dependencies through Transformer, and improving classification robustness through the dynamic routing mechanism of the capsule network, the method effectively improves the accuracy and generalization ability of EEG emotion recognition.

This invention relates to the field of computer vision, specifically to a passive domain adaptive target detection method and electronic device based on pseudo-label confidence feedback, comprising the following steps: S1, acquiring a pre-trained source domain model and unlabeled target domain data; S2, constructing a mean-teacher self-training architecture; S3, generating pseudo-labels and calculating confidence feedback signals; S4, implementing an adaptive dynamic mean-teacher update strategy. This invention, by introducing a pseudo-label confidence feedback mechanism, enables the model to have adaptive adjustment capabilities, effectively suppressing the adverse effects of low-quality pseudo-labels on model training; it can dynamically reduce the decay rate in high-quality pseudo-label scenarios, improving cross-domain transfer efficiency and detection accuracy; this invention effectively alleviates the "confirmation bias" problem in passive domain adaptive tasks, significantly improving the model's training stability, noise resistance, and cross-domain target detection performance in the target domain, possessing good application value and promising prospects for promotion.

The present application relates to the technical field of aluminum electrolysis, and provides an aluminum electrolysis cell damage inversion method based on a machine learning algorithm, comprising the following steps: obtaining current data of an aluminum electrolysis cell through a fiber-optic current sensor, and obtaining temperature data of the aluminum electrolysis cell; performing discharge area division according to a discharger of the aluminum electrolysis cell; performing area division on the current data and the temperature data, and performing feature extraction on the current data and the temperature data; establishing a physical information forward model based on a neural network, and associating and mapping a damage state vector with current feature data and temperature feature data through the physical information forward model; constructing a cathode joint loss function and an anode joint loss function; respectively performing iterative solution on the cathode joint loss function and the anode joint loss function through an optimization algorithm, outputting an optimal damage state vector, and generating a damage space distribution result of the electrolysis cell. The present application can realize online inversion and spatial distribution visualization of anode and cathode damage states.

The application provides a lightweight millimeter wave radar two-dimensional feature map classification method and system for FPGA hardware deployment, and solves the technical problems of poor real-time performance and high power consumption of existing millimeter wave radar two-dimensional feature map classification methods in edge computing scenarios. It includes obtaining a feature map and preprocessing it to obtain a preprocessed feature map, constructing a dataset, and dividing the dataset into a training set and a validation set; building a classification network model based on LeNet, training the classification network model using the training set, and obtaining a trained classification network model; verifying whether the performance of the trained classification network model meets the performance indicators using the validation set, and if so, obtaining a validated classification network model, deploying it to an FPGA, classifying the feature map based on the classification network model, and obtaining a classification result; otherwise, continue training to update the parameters of the classification network model until the performance after verification meets the performance indicators. The application can be widely applied in the field of image classification technology.

The invention discloses a porosity prediction method and device for a clastic rock reservoir, equipment and a medium, and relates to the technical field of oil and gas reservoir prediction, data cleaning, abnormal value detection and processing and feature engineering processing are performed on logging effective sample data of the clastic rock reservoir, and a sample data set is obtained; constructing a porosity prediction model based on a deep learning algorithm and an integrated learning algorithm; dividing the sample data set by using a weighted integrated multi-algorithm framework, and performing data preprocessing on the divided data set to obtain a target data set; performing model training on the porosity prediction model based on a programming language framework, and performing model testing and performance evaluation on the trained porosity prediction model; taking the porosity prediction model passing the performance evaluation as a target porosity prediction model; and inputting the target well section data into the target porosity prediction model, and outputting a porosity prediction value, thereby improving the precision and stability of porosity prediction, and solving the problem that a complex nonlinear relationship cannot be effectively processed.

The invention relates to the crossing field of public health and geographic information technology, and discloses a plague risk assessment method and system combined with a MaxEnt model, and the method comprises the steps: obtaining a multi-source environment variable, and carrying out the standardization; reducing a colinear variable through variance threshold preliminary screening, a Pearson's correlation coefficient and plague point significance test; performing variable importance sorting by using recursive feature elimination and a support vector machine, and dynamically determining an optimal variable subset; and inputting a MaxEnt model to train ecological niche probability distribution, and dividing risk levels based on historical occurrence point probability quantiles. The system comprises corresponding function modules. According to the system, through a three-stage automatic variable screening mechanism, the generalization ability, interpretability and prediction precision of the model are remarkably improved, and reliable support is provided for accurate prevention and control of plague.

The application discloses a plunger pump wear fault detection method based on pressure-temperature layered triggering and physical prior mask self-attention, and belongs to the technical field of hydraulic system state monitoring and fault diagnosis. The method comprises the following steps: collecting vibration, pressure and temperature multi-source signals, and setting a default low power consumption of a main controller; realizing layered early warning and deep diagnosis awakening based on four dynamic characteristics of pressure and temperature and an abnormal proportion of a sliding window; performing synchronous compression wavelet transform on the vibration signal and performing temperature-pressure analytical linear real-time correction, block encoding and vibration Token sequence generation; generating an additive physical mask guided self-attention calculation based on a fault frequency band prior, and extracting wear sensitive features; comparing with a working condition grouping fault feature library to determine the fault type and grade. The application runs through the whole chain of signal correction, feature fusion and diagnosis decision with physical knowledge, and under the lightweight design of a model parameter quantity less than 500k, realizes a variable working condition diagnosis accuracy of 94.2% and an early wear false negative rate of 3.5%.