zggg1p / minist-dataset-classification Goto Github PK

This task focuses on training the MINIST dataset to get a feel for the differences between fully connected networks, neural networks, and the effects of different activation functions, training techniques (BN, Dropout, and L2 regularization) on the network. 本次任务主要是通过对MINIST数据集进行训练来感受全连接网络、神经网络的差异，以及不同激活函数、训练技巧（BN，Dropout以及L2正则化）对网络的影响。