LGDec 22, 2023

How Smooth Is Attention?

arXiv:2312.14820v239 citationsh-index: 19ICML
Originality Highly original
AI Analysis

This provides foundational insights into the robustness and expressive power of attention mechanisms, which is crucial for the broader ML/AI community working with Transformers.

The paper tackled the problem of mathematically understanding the Lipschitz properties of self-attention in Transformers, showing that the Lipschitz constant is bounded by √n up to a constant factor for reasonable sequence lengths and providing independent bounds in a mean-field regime.

Self-attention and masked self-attention are at the heart of Transformers' outstanding success. Still, our mathematical understanding of attention, in particular of its Lipschitz properties - which are key when it comes to analyzing robustness and expressive power - is incomplete. We provide a detailed study of the Lipschitz constant of self-attention in several practical scenarios, discussing the impact of the sequence length $n$ and layer normalization on the local Lipschitz constant of both unmasked and masked self-attention. In particular, we show that for inputs of length $n$ in any compact set, the Lipschitz constant of self-attention is bounded by $\sqrt{n}$ up to a constant factor and that this bound is tight for reasonable sequence lengths. When the sequence length $n$ is too large for the previous bound to be tight, which we refer to as the mean-field regime, we provide an upper bound and a matching lower bound which are independent of $n$. Our mean-field framework for masked self-attention is novel and of independent interest. Our experiments on pretrained and randomly initialized BERT and GPT-2 support our theoretical findings.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes