Dynamic Neural Networks: A Survey

TL;DR

This survey reviews the rapidly evolving field of dynamic neural networks, which adapt their structures or parameters based on input data, offering advantages in accuracy and efficiency over static models.

Contribution

It categorizes dynamic neural networks into instance-wise, spatial-wise, and temporal-wise types, systematically reviewing key research problems and future directions.

Findings

Dynamic networks improve accuracy and efficiency.

Categorization into three main types of dynamic models.

Identification of open problems and future research directions.

Abstract

Dynamic neural network is an emerging research topic in deep learning. Compared to static models which have fixed computational graphs and parameters at the inference stage, dynamic networks can adapt their structures or parameters to different inputs, leading to notable advantages in terms of accuracy, computational efficiency, adaptiveness, etc. In this survey, we comprehensively review this rapidly developing area by dividing dynamic networks into three main categories: 1) instance-wise dynamic models that process each instance with data-dependent architectures or parameters; 2) spatial-wise dynamic networks that conduct adaptive computation with respect to different spatial locations of image data and 3) temporal-wise dynamic models that perform adaptive inference along the temporal dimension for sequential data such as videos and texts. The important research problems of dynamic networks, e.g., architecture design, decision making scheme, optimization technique and applications, are reviewed systematically. Finally, we discuss the open problems in this field together with interesting future research directions.