8results about How to "Enhance aggregation ability" patented technology

The application discloses a federated learning method and system based on modulus component homomorphic encryption and response verification, relates to the technical field of information security, and comprises the following steps: obtaining initial global model parameters and generating a pre-shared key; a client generates a participation key and a calculation key; a plurality of clients are randomly selected as participation clients and a challenge pair is sent; a participation client generates a challenge response value based on the pre-shared key and the challenge pair, performs local training based on the initial global model parameters, encrypts local model parameters based on the participation key and a modulus component homomorphic encryption algorithm, and obtains model parameter ciphertext; trusted verification is performed based on the challenge response value, and encrypted global model parameters are obtained by aggregating trusted model parameter ciphertext based on the calculation key; the participation client updates the local model based on the encrypted global model parameters, the above process is repeated until the model converges, and a trained global model is obtained. The encryption calculation efficiency, response speed and security robustness of the federated learning are improved.

This invention discloses a lightweight image super-resolution reconstruction method, system, and storage medium, relating to the field of image processing technology. Specifically, it includes: acquiring high-resolution and low-resolution images of the target as training sample pairs; constructing a lightweight image super-resolution reconstruction model comprising a shallow feature extraction module, a deep feature extraction module, a multi-layer feature fusion module, and an image reconstruction module. The shallow feature extraction module performs shallow processing on the low-resolution image; the deep feature extraction module consists of multiple cascaded enhanced separable residual feature refinement blocks, extracting shallow features to obtain multiple deep features; the multi-layer feature fusion module is used to obtain fused features; and the image reconstruction module is used to output the high-resolution image. The trained model is used to reconstruct the low-resolution image to be processed. This invention has a simple network structure and achieves accurate reconstruction of image texture and details using only a low number of parameters and computational cost.

The application provides a foreground-guided cross-scale memory attention self-supervised monocular depth estimation method and system, and relates to the technical fields of computer vision, three-dimensional visual perception and self-supervised depth estimation.The foreground-guided self-supervised monocular depth estimation network is constructed, and an encoder, a self-generated foreground prior branch, a foreground-guided cross-scale memory grouping query attention bottleneck module and a depth decoder are used as main construction modules.In the self-supervised monocular depth estimation task, the self-generated foreground prior branch is used to learn to predict the foreground prior under the supervision of the offline generated binary pseudo foreground mask, so that the network can generate the binary foreground prior according to the single target image in the inference stage without an external segmentation model.The foreground-guided cross-scale memory grouping query attention bottleneck module is used to fuse deep layer encoding features and intermediate layer encoding features, construct cross-scale shared key-value memories, and construct a foreground selective attention bias according to the binary foreground prior, so as to enhance the information aggregation in the foreground region, and at the same time, not to explicitly suppress the attention path between the foreground and the background, thereby improving the depth prediction stability and integrity of the moving foreground region, the occlusion region and the depth discontinuous boundary.

The invention discloses an alarm data aggregation method and device, equipment and a medium, and belongs to the technical field of computers. According to the method, the importance of condition attributes relative to each equivalence class corresponding to a preset decision attribute set is introduced, and according to the difference between the importance of each condition attribute in the preset condition attribute set relative to the equivalence class and the importance of the whole preset condition attribute set relative to the equivalence class, the equivalence class is determined. And determining an attribute weight of each condition attribute corresponding to the equivalence class, thereby identifying important attributes in respective associated condition attributes under different equivalence classes, and further judging whether the related alarm data is associated with the same attack classification according to the important attributes under each equivalence class. And the alarm data associated with the same attack classification is aggregated.

The present application relates to the field of olefin polymerization catalyst, and provides a metal-silane composite catalyst and a preparation method and application thereof.The metal-silane composite catalyst comprises a mixture of diazabicycloalkyldialkoxy silane compounds, organic compounds containing C-O and / or C=O, metal halides and / or reaction products of the above components, and further comprises superfine inorganic oxide; after mixing the metal-silane composite catalyst and organic aluminum compound, the mixture can be used for olefin polymerization reaction.The polymer prepared by using the catalyst has a high melt index and low oligomer content.

The application relates to the technical field of artificial intelligence, in particular to a multi-scale causal shunt prediction method and system for an online soft measurement of a cut tobacco process. Multivariate process data and outlet material moisture data of the cut tobacco process are acquired; the data is subjected to time synchronization, missing value processing, abnormal value processing and standardization preprocessing, and a sliding time window sample is constructed; input features are divided into local path features and global path features along a channel dimension, the local path features are subjected to local mapping, and the global path features are subjected to attention calculation with a causal mask; the shunt modeling result is input into a multi-scale causal coding backbone network to extract hierarchical features, and the different hierarchical features are subjected to time alignment and cross-layer fusion; local transient disturbance and high-frequency change information are extracted through a high-frequency residual branch; the main road fusion features and the high-frequency compensation features are jointly input into a prediction head to output an outlet material moisture prediction result. The application aims to solve the problem of how to improve the online soft measurement precision of the cut tobacco process and reduce the calculation overhead.

The present application discloses a metal complex containing silicon-containing heterofluorene group, an olefin polymerization catalyst containing the metal complex, and a preparation method and application thereof. The metal complex contains a main complex which is a metal compound containing silicon-containing heterofluorene group, the metal compound is a mixture of metal halide, dialkyl silicon-containing heterofluorene compound, electron donor compound and / or reaction product of the above components, and the silicon-containing heterofluorene group contained therein makes the catalyst containing the same have higher activity. The above main complex is loaded on an inorganic oxide carrier with a specific particle size range to obtain a high-dispersity catalyst, and the catalyst can be used in olefin polymerization reaction, and the polymer obtained by polymerization has higher melt index and lower oligomer content.

The application discloses a medical time sequence classification method based on time-aware multi-instance learning, and belongs to the technical field of signal processing. The method comprises the following steps: acquiring a medical time sequence collected by a wearable sensor; and inputting the medical time sequence into a medical time sequence classification model, wherein the medical time sequence classification model comprises a time-aware module, a multi-granularity block embedding module, a multi-granularity attention module and a classification output unit. The method is based on time sequence input, sequentially processed through the time-aware module, the multi-granularity block embedding module and the multi-granularity attention module, and finally outputs a classification label result, so that automatic evaluation of Parkinson's disease is realized.