Bespoke vs. Prêt-à-Porter Lottery Tickets: Exploiting Mask Similarity for Trainable Sub-Network Finding
This work addresses the problem of identifying reliable trainable sub-networks in neural networks for researchers, showing incremental improvements in LT generation.
The study investigated the uniqueness and connectivity of Lottery Tickets (LTs) across 28 image classification tasks and architectures, finding differences in mask structure based on pruning methods and disproving uniqueness. It proposed a consensus-based denoising method to refine LTs, achieving comparable performance to ordinary LTs without additional pruning iterations.
The observation of sparse trainable sub-networks within over-parametrized networks - also known as Lottery Tickets (LTs) - has prompted inquiries around their trainability, scaling, uniqueness, and generalization properties. Across 28 combinations of image classification tasks and architectures, we discover differences in the connectivity structure of LTs found through different iterative pruning techniques, thus disproving their uniqueness and connecting emergent mask structure to the choice of pruning. In addition, we propose a consensus-based method for generating refined lottery tickets. This lottery ticket denoising procedure, based on the principle that parameters that always go unpruned across different tasks more reliably identify important sub-networks, is capable of selecting a meaningful portion of the architecture in an embarrassingly parallel way, while quickly discarding extra parameters without the need for further pruning iterations. We successfully train these sub-networks to performance comparable to that of ordinary lottery tickets.