PyTorch - Dropout: A Simple Way to Prevent Neural Networks from Overfitting

interview ：

• dropout What is the principle and function of ？
• dropout How to solve the problem of over fitting , What kind of learning methods are simulated ？ answer ： Integrated learning
• stay dropout in , In addition to directly weight Set to 0, For reserved weight What else to do , for example p=0.2
• Achieve one dropout() function , Input input, Output output, Three layer parameters input,p,training
• call model.eval(), What to do , And before calling , What's the difference ？ Which are commonly used layer Affected ？

torch.nn Module , Contains a lot of layer, Neural network is the core module , All are torch.nn The combination of

combination torch.nn Source code , Understanding different API, It also includes classic papers

Dropout Layers, besides , In the form of a function dropout,torch.nn.functional.dropout

• nn.Dropout

torch.nn.Dropout(p=0.5, inplace=False)

import</