Dropout

Introduced by Srivastava et al. in Dropout: A Simple Way to Prevent Neural Networks from Overfitting

Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is $p=0.5$). At test time, all units are present, but with weights scaled by $p$ (i.e. $w$ becomes $pw$).

The idea is to prevent co-adaptation, where the neural network becomes too reliant on particular connections, as this could be symptomatic of overfitting. Intuitively, dropout can be thought of as creating an implicit ensemble of neural networks.

Source: Dropout: A Simple Way to Prevent Neural Networks from Overfitting

Read Paper See Code

Papers

Paper	Code	Results	Date	Stars

Tasks

Task	Papers	Share
Language Modelling	51	6.50%
Retrieval	30	3.82%
Question Answering	27	3.44%
Large Language Model	26	3.31%
Semantic Segmentation	22	2.80%
In-Context Learning	15	1.91%
Object Detection	14	1.78%
Image Classification	12	1.53%
Sentence	11	1.40%

Usage Over Time

This feature is experimental; we are continuously improving our matching algorithm.

Components

Component	Type	Add Remove
🤖 No Components Found	You can add them if they exist; e.g. Mask R-CNN uses RoIAlign

Categories

Add Remove

Regularization