首页|Study Results from Beijing University of Posts and Telecommunications Broaden Un derstanding of Artificial Intelligence (Adaptive negative representations for gr aph contrastive learning)
Study Results from Beijing University of Posts and Telecommunications Broaden Un derstanding of Artificial Intelligence (Adaptive negative representations for gr aph contrastive learning)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Current study results on artificial in telligence have been published. According to news reporting out of Beijing, Peop le's Republic of China, by NewsRx editors, research stated, "Graph contrastive l earning (GCL) has emerged as a promising paradigm for learning graph representat ions." Financial supporters for this research include National Natural Science Foundati on of China; National Key Research And Development Program of China. Our news editors obtained a quote from the research from Beijing University of P osts and Telecommunications: "Recently, the idea of hard negatives is introduced to GCL, which can provide more challenging self-supervised objectives and allev iate over-fitting issues. These methods use different graphs in the same mini-ba tch as negative examples, and assign larger weights to true hard negative ones. However, the influence of such weighting strategies is limited in practice, sinc e a small mini-batch may not contain any challenging enough negative examples. I n this paper, we aim to offer a more flexible solution to affect the hardness of negatives by directly manipulating the representations of negatives. By assumin g that (1) good negative representations should not deviate far from the represe ntations of real graph samples, and (2) the computation process of graph encoder may introduce biases to graph representations, we first design a negative repre sentation generator (NRG) which (1) employs real graphs as prototypes to perturb , and (2) introduces parameterized perturbations through the feed-forward comput ation of the graph encoder to match the biases. Then we design a generation loss to train the parameters in NRG and adaptively generate negative representations for more challenging contrastive objectives."
Beijing University of Posts and Telecomm unicationsBeijingPeople's Republic of ChinaAsiaArtificial Intelligence
NETL
NSTL
