Dropout: a simple way to prevent neural networks from overfitting

TLDR

Deep neural nets with many parameters are powerful but prone to overfitting, and their size makes ensemble approaches impractical. Dropout is proposed as a simple technique to mitigate overfitting in large neural networks. During training, dropout randomly removes units and connections, sampling from an exponential number of thinned networks, and at test time a single network with scaled weights approximates the ensemble average. Dropout substantially reduces overfitting and outperforms other regularization methods, achieving state‑of‑the‑art results on vision, speech, document classification, and computational biology benchmarks.

Abstract

Deep neural nets with a large number of parameters are very powerful machine learning systems. However, overfitting is a serious problem in such networks. Large networks are also slow to use, making it difficult to deal with overfitting by combining the predictions of many different large neural nets at test time. Dropout is a technique for addressing this problem. The key idea is to randomly drop units (along with their connections) from the neural network during training. This prevents units from co-adapting too much. During training, dropout samples from an exponential number of different networks. At test time, it is easy to approximate the effect of averaging the predictions of all these thinned networks by simply using a single unthinned network that has smaller weights. This significantly reduces overfitting and gives major improvements over other regularization methods. We show that dropout improves the performance of neural networks on supervised learning tasks in vision, speech recognition, document classification and computational biology, obtaining state-of-the-art results on many benchmark data sets.