Class-Incremental Learning for Semantic Segmentation in Aerial Imagery via Distillation in All Aspects

Graph convolutional neural network for individual consumption behavior analysis

“…In our method, sessions are represented as graphs and GNN is used to learn object representation. In addition to using the attention mechanism, our method also makes up for the deficiency of RNN [13] in learning object representation to some extent, making its performance better than the previous model. It is important to note that our approach introduces a repeated-exploration mechanism that can predict results more accurately than other models that do not use Transformer.…”

Multi-level fast personalized recommendation system

“…Because of its powerful feature representation ability, it has received great attention in the fields such as computer vision and natural language processing [6,7,8]. Text, images, and videos are all data defined on a regular grid, which can be viewed as distributed on one -, two -, and three-dimensional grid support sets [9,10,11,12,13]. Graph convolutional networks have attracted more and more attention by using their powerful feature representation ability to improve learning effects.…”

Facial expression recognition combining different resolution images

“…Secondly, in recent years, feedforward neural networks (FNN) and convolutional neural networks (CNN) have also been used to extract facial features. A new recognition framework based on convolutional neural networks (CNN) has achieved remarkable results in FER [1][2][3][4][5]. Multiple convolution and aggregation layers in CNN can extract higher and multi-level features of the whole face or local area and has a good classification performance of facial expression image features.…”