Exploring Explicit Coarse-Grained Structure in Artificial Neural Networks

作     者：杨析辞 谢志远 杨晓涛 Xi-Ci Yang;Z.Y.Xie;Xiao-Tao Yang

作者机构：College of Power and Energy EngineeringHarbin Engineering UniversityHarbin 150001China Department of PhysicsRenmin University of ChinaBeijing 100872China 

出 版 物：《Chinese Physics Letters》 (中国物理快报（英文版）)

年 卷 期：2023年第40卷第2期

页      面：6-14页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported by the National R&D Program of China(Grant Nos.2017YFA0302900 and 2016YFA0300503) the National Natural Science Foundation of China(Grant Nos.52176064,12274458,and 11774420) the Research Funds of Renmin University of China(Grant No.20XNLG19) 

主　　题：neural artificial hierarchical 

摘      要：We propose to employ a hierarchical coarse-grained structure in artificial neural networks explicitly to improve the interpretability without degrading *** idea has been applied in two *** is a neural network called Taylor Net,which aims to approximate the general mapping from input data to output result in terms of Taylor series directly,without resorting to any magic nonlinear *** other is a new setup for data distillation,which can perform multi-level abstraction of the input dataset and generate new data that possesses the relevant features of the original dataset and can be used as references for *** both the cases,the coarse-grained structure plays an important role in simplifying the network and improving both the interpretability and *** validity has been demonstrated on MNIST and CIFAR-10 *** improvement and some open questions related are also discussed.