A contrastive rule for meta-learning

Official implementation of the paper A contrastive rule for meta-learning published at NeurIPS 2022.

Usage

• metaopt_spiking/ implements the meta-optimization experiments of section 5.2 and the recurrent spiking network experiments of section 5.4
• fewshot/ implements the visual few-shot experiments of section 5.3
• bandit/ implements the reward-based learning experiment of section 5.5

Dependencies

The meta-optimization (section 5.2), visual few-shot learning (section 5.3) and recurrent spiking network (section 5.4) experiments are implemented using pytorch, the reward-based learning experiment (section 5.5) is implemented using jax.

For specific package dependencies see the respective subfolder's requirements.txt files.