Sparse Attentive Backtracking: Temporal CreditAssignment Through Reminding

TL;DR

Sparse Attentive Backtracking introduces a biologically inspired method for learning long-term dependencies in sequences by using attention-based memory associations, reducing computational costs compared to traditional backpropagation methods.

Contribution

The paper proposes a novel attention-based algorithm that enables credit assignment over long sequences without full backpropagation through time, inspired by human memory recall.

Findings

Outperforms BPTT and truncated BPTT on long-term dependency tasks.

Transfers better to longer sequences than LSTMs trained with BPTT or full self-attention.

Requires fewer backward passes, reducing computational complexity.

Abstract

Learning long-term dependencies in extended temporal sequences requires credit assignment to events far back in the past. The most common method for training recurrent neural networks, back-propagation through time (BPTT), requires credit information to be propagated backwards through every single step of the forward computation, potentially over thousands or millions of time steps. This becomes computationally expensive or even infeasible when used with long sequences. Importantly, biological brains are unlikely to perform such detailed reverse replay over very long sequences of internal states (consider days, months, or years.) However, humans are often reminded of past memories or mental states which are associated with the current mental state. We consider the hypothesis that such memory associations between past and present could be used for credit assignment through arbitrarily long sequences, propagating the credit assigned to the current state to the associated past state. Based on this principle, we study a novel algorithm which only back-propagates through a few of these temporal skip connections, realized by a learned attention mechanism that associates current states with relevant past states. We demonstrate in experiments that our method matches or outperforms regular BPTT and truncated BPTT in tasks involving particularly long-term dependencies, but without requiring the biologically implausible backward replay through the whole history of states. Additionally, we demonstrate that the proposed method transfers to longer sequences significantly better than LSTMs trained with BPTT and LSTMs trained with full self-attention.