首页|Reports from Soochow University Advance Knowledge in Machine Learning (Semi-supe rvised Class Incremental Broad Network for Continuous Diagnosis of Rotating Mach inery Faults With Limited Labeled Samples)
Reports from Soochow University Advance Knowledge in Machine Learning (Semi-supe rvised Class Incremental Broad Network for Continuous Diagnosis of Rotating Mach inery Faults With Limited Labeled Samples)
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-Researchers detail new data in Machine Learning. According to news originating from Suzhou, People's Republic of China , by NewsRx correspondents, research stated, "Machine learning has demonstrated remarkable success in the field of intelligent fault diagnosis. In current machi ne learning systems, models tend to be fixed after training, which makes them ca n only generalize to classes that appear in the training set, and cannot continu ously learn newly emerging classes." Financial supporters for this research include National Natural Science Foundati on of China (NSFC), Natural Science Foundation of Jiangsu Province. Our news journalists obtained a quote from the research from Soochow University, "However, in real industrial scenarios, industrial data is constantly growing, which requires models to be able to acquire new knowledge while retaining existi ng knowledge. Furthermore, a substantial proportion of the acquired samples in i ndustry are unlabeled and only few samples are labeled. To tackle the above two challenges, this article proposes a novel semi-supervised class incremental broa d network (SSCIBN) for incremental intelligent diagnosis of rotating machinery f aults under limited labeled samples. Specifically, a semisupervised graph embed ding loss function is designed. On the one hand, this loss function can learn th e structural information of a large amount of unlabeled data, which overcomes th e limitation that the scarcity of labeled samples leads to poor model performanc e. On the other hand, local structure information within old and new classes is considered in the continuous learning process to fully learn the unique manifold structure information of different classes, which in turn enhances the discrimi native performance between old and new classes. Furthermore, a novel semisupervi sed class incremental learning mechanism is proposed, which does not need to uti lize the old class data during the incremental learning process, but can effecti vely retain the old class knowledge. The effectiveness of the proposed method is evaluated through multiple mechanical failure increment cases."
SuzhouPeople's Republic of ChinaAsiaCyborgsEmerging TechnologiesMachine LearningSoochow University
NETL
NSTL
