Abstract: The present invention belongs to the technical field of data security of industrial Internet, and provides a method and system for detecting abnormal nodes in industrial Internet, a medium and a device. In the industrial Internet, different data holders firstly transform their own local node data into graphic data. Before local model training, the data holders firstly use a spectral clustering algorithm to perform certain clustering operations on local data, cluster the node data of the same category into the same cluster, and then perform local model training on a clustered result to obtain partial aggregation features. The trained partial features are uploaded to a trusted third-party server for global feature aggregation. Through an attention mechanism, different weights are assigned for partial features uploaded by different data holders, and the aggregated global features are delivered to each data holder for a new round of training.

Abstract: The present invention belongs to the technical field of data security of industrial Internet, and provides a method and system for detecting abnormal nodes in industrial Internet, a medium and a device. In the industrial Internet, different data holders firstly transform their own local node data into graphic data. Before local model training, the data holders firstly use a spectral clustering algorithm to perform certain clustering operations on local data, cluster the node data of the same category into the same cluster, and then perform local model training on a clustered result to obtain partial aggregation features. The trained partial features are uploaded to a trusted third-party server for global feature aggregation. Through an attention mechanism, different weights are assigned for partial features uploaded by different data holders, and the aggregated global features are delivered to each data holder for a new round of training.

Abstract: The present invention relates to the technical field of expression recognition, and in particular, to an expression recognition method and system based on multi-scale features and spatial attention. The method includes: performing feature extraction on acquired facial image data by using an HNFER neural network model to obtain an original input feature map; performing pooling and concatenation on extracted features based on a CoordAtt attention mechanism to obtain a feature map; performing deep convolution processing on the feature map to obtain an attention map, and then performing element-by-element multiplication to obtain a final feature map; and performing feature transformation and normalization on the final feature map to obtain an expression category probability and output the expression category probability.