Going Beyond Linear Transformers with Recurrent Fast Weight Programmers

Kazuki Irie, Imanol Schlag, Róbert Csordás, Jürgen Schmidhuber

Keywords: Transformers, memory augmented recurrent neural networks, fast weight programmers

Abstract: Transformers with linearised attention (''linear Transformers'') have demonstrated the practical scalability and effectiveness of outer product-based Fast Weight Programmers (FWPs) from the '90s. However, the original FWP formulation is more general than the one of linear Transformers: a slow neural network (NN) continually reprograms the weights of a fast NN with arbitrary architecture. In existing linear Transformers, both NNs are feedforward and consist of a single layer. Here we explore new variations by adding recurrence to the slow and fast nets. We evaluate our novel recurrent FWPs (RFWPs) on two synthetic algorithmic tasks (code execution and sequential ListOps), Wikitext-103 language models, and on the Atari 2600 2D game environment. Our models exhibit properties of Transformers and RNNs. In the reinforcement learning setting, we report large improvements over LSTM in several Atari games. Our code is public.

Code Of Conduct: I certify that all co-authors of this work have read and commit to adhering to the NeurIPS Statement on Ethics, Fairness, Inclusivity, and Code of Conduct.

TL;DR: We augment linear Transformers with recurrent connections.

Code: https://github.com/IDSIA/recurrent-fwp

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2106.06295/code)