Code for our work "Autoequivariant Network Search via Group Decomposition": https://arxiv.org/abs/2104.04848 We show that larger groups can be broken down into smaller ones to induce group equivariance efficiently. Further, we use reinforcement learning to automatically induce the correct form of equivariance in any dataset. The codes are organized into the following folders: - MNIST Single Equivariance Test: Performs single equivariance test for various augmented MNIST augmented datasets and equivariances - FashionMNIST Single Equivariance Test: Performs single equivariance test for various augmented FashionMNIST augmented datasets and equivariances - MNIST DQN: Deep-Q learning for various augmented MNIST - FashionMNIST DQN: Deep-Q learning for various augmented FashionMNIST - DQN-GCNN: Deep-Q learning for various image datasets using GCNN More details are provided in individual readme files in each folders