Dropout

• Dropout means dropping out some of the nodes of the neural network to avoid Overfitting
  • This is done randomly based on given probability
• Dropout can be thought of ensemble of multiple networks
• One way to solve Overfitting is ensemble of multiple neural networks
  • But for a big neural network, it is inefficient and time consuming
• So one of the way is to use Dropout.
  • Dropout randomly drops out nodes on the training time from the parent network
  • Hence create different models for each batch