Sparse Attentive Backtracking: Long-Range Credit Assignment in Recurrent Networks
This addresses a major bottleneck in training recurrent neural networks on long sequences, potentially improving efficiency and biological plausibility, though it appears incremental as an enhancement to existing gradient-based methods.
The paper tackles the problem of learning long-term dependencies in recurrent networks by addressing the computational and biological limitations of backpropagation through time (BPTT), proposing Sparse Attentive Backtracking as an alternative that uses attention to selectively backpropagate through relevant past states, resulting in efficient learning without truncation biases.
A major drawback of backpropagation through time (BPTT) is the difficulty of learning long-term dependencies, coming from having to propagate credit information backwards through every single step of the forward computation. This makes BPTT both computationally impractical and biologically implausible. For this reason, full backpropagation through time is rarely used on long sequences, and truncated backpropagation through time is used as a heuristic. However, this usually leads to biased estimates of the gradient in which longer term dependencies are ignored. Addressing this issue, we propose an alternative algorithm, Sparse Attentive Backtracking, which might also be related to principles used by brains to learn long-term dependencies. Sparse Attentive Backtracking learns an attention mechanism over the hidden states of the past and selectively backpropagates through paths with high attention weights. This allows the model to learn long term dependencies while only backtracking for a small number of time steps, not just from the recent past but also from attended relevant past states.