# ShuffleNASNets: Efficient CNN models through modified Efficient Neural Architecture Search

## TL;DR

ShuffleNASNets leverage expert knowledge within a limited search space to produce efficient CNN models that are faster, less complex, and maintain competitive accuracy on CIFAR-10, including low-parameter variants.

## Contribution

Integrating expert design principles into ENAS to create ShuffleNASNets that are more efficient and scalable while maintaining high accuracy.

## Key findings

- ShuffleNASNets are twice as fast as ENAS baseline.
- They require fewer parameters and are less complex.
- Achieve less than 5% test error with only 236K parameters.

## Abstract

Neural network architectures found by sophistic search algorithms achieve strikingly good test performance, surpassing most human-crafted network models by significant margins. Although computationally efficient, their design is often very complex, impairing execution speed. Additionally, finding models outside of the search space is not possible by design. While our space is still limited, we implement undiscoverable expert knowledge into the economic search algorithm Efficient Neural Architecture Search (ENAS), guided by the design principles and architecture of ShuffleNet V2. While maintaining baseline-like 2.85% test error on CIFAR-10, our ShuffleNASNets are significantly less complex, require fewer parameters, and are two times faster than the ENAS baseline in a classification task. These models also scale well to a low parameter space, achieving less than 5% test error with little regularization and only 236K parameters.