首页|Findings from China Iron and Steel Research Institute GroupBroaden Understandin g of Machine Learning (Prediction of OpticalProperties of Oxide Glass Combined With Autoencoder andMachine Learning)
Findings from China Iron and Steel Research Institute GroupBroaden Understandin g of Machine Learning (Prediction of OpticalProperties of Oxide Glass Combined With Autoencoder andMachine 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 DailyNews Daily News – Fresh data on Machine Learning are pre sented in a new report. According to newsoriginating from Beijing, People’s Rep ublic of China, by NewsRx correspondents, research stated, “Thecomposition of o xide glasses is characterized by high dimensionality and sparsity, making it cha llengingto establish high-precision predictive models. Therefore, feature extra ction is essential.”Financial support for this research came from National Key Research & Development Program of China.Our news journalists obtained a quote from the research from China Iron and Stee l Research InstituteGroup, “This study focuses on the optical properties of oxi de glasses (refractive index and Abbe number),utilizing autoencoder (AE) and ma chine learning techniques to achieve automated feature extraction.The results i ndicate that compared to standalone neural networks (NN), AE-NN transforms unsup ervisedlearning into supervised learning, reducing feature dimensions while imp roving model accuracy. Specifically,for the refractive index dataset, the dimen sionality was reduced from 63 to 25, with a corresponding testset coefficient o f determination (R2) of 0.95. For the Abbe number dataset, the dimensionality wa sreduced from 61 to 30, with a corresponding test set R2 of 0.97, demonstrating the effectiveness of thefeature extraction method. Regarding interpretability, analyzing the encoder weight matrix of the AE-NNidentified the importance of o riginal features, with Co and Y being the most significant for both refractivei ndex and Abbe number. Additionally, the application of the feature extraction me thod in machinelearning models shows its generality in improving model performa nce, particularly for nonensemble modelssuch as Support Vector Regression (SVR) or k-Nearest Neighbors (KNN), exhibiting significant accuracyenhancements.”
BeijingPeople’s Republic of ChinaAsi aCyborgsEmergingTechnologiesMachine LearningChina Iron and Steel Resear ch Institute Group
NETL
NSTL
