Playing the lottery with rewards and multiple languages: lottery tickets in RL and NLP

TL;DR

This paper demonstrates that the lottery ticket hypothesis applies broadly across NLP and RL, showing that smaller, pruned models initialized with winning tickets can perform comparably to larger models, indicating a general phenomenon in deep learning.

Contribution

The study extends the lottery ticket hypothesis to NLP and RL, showing winning tickets exist in these domains and enable significant model size reduction without performance loss.

Findings

Winning tickets outperform random initializations at high pruning rates

Transformers can be pruned to one-third size with similar performance

Lottery ticket phenomenon applies beyond supervised image classification

Abstract

The lottery ticket hypothesis proposes that over-parameterization of deep neural networks (DNNs) aids training by increasing the probability of a "lucky" sub-network initialization being present rather than by helping the optimization process (Frankle & Carbin, 2019). Intriguingly, this phenomenon suggests that initialization strategies for DNNs can be improved substantially, but the lottery ticket hypothesis has only previously been tested in the context of supervised learning for natural image tasks. Here, we evaluate whether "winning ticket" initializations exist in two different domains: natural language processing (NLP) and reinforcement learning (RL).For NLP, we examined both recurrent LSTM models and large-scale Transformer models (Vaswani et al., 2017). For RL, we analyzed a number of discrete-action space tasks, including both classic control and pixel control. Consistent with workin supervised image classification, we confirm that winning ticket initializations generally outperform parameter-matched random initializations, even at extreme pruning rates for both NLP and RL. Notably, we are able to find winning ticket initializations for Transformers which enable models one-third the size to achieve nearly equivalent performance. Together, these results suggest that the lottery ticket hypothesis is not restricted to supervised learning of natural images, but rather represents a broader phenomenon in DNNs.