Near-Optimal Glimpse Sequences for Improved Hard Attention Neural Network Training
This work addresses computational efficiency issues in computer vision for researchers and practitioners, but it is incremental as it builds on existing BOED methods.
The paper tackles the problem of high-variance training in hard visual attention mechanisms for image classification by framing it as a Bayesian optimal experimental design problem, resulting in near-optimal attention sequences that speed up training, though generation is computationally expensive.
Hard visual attention is a promising approach to reduce the computational burden of modern computer vision methodologies. Hard attention mechanisms are typically non-differentiable. They can be trained with reinforcement learning but the high-variance training this entails hinders more widespread application. We show how hard attention for image classification can be framed as a Bayesian optimal experimental design (BOED) problem. From this perspective, the optimal locations to attend to are those which provide the greatest expected reduction in the entropy of the classification distribution. We introduce methodology from the BOED literature to approximate this optimal behaviour, and use it to generate `near-optimal' sequences of attention locations. We then show how to use such sequences to partially supervise, and therefore speed up, the training of a hard attention mechanism. Although generating these sequences is computationally expensive, they can be reused by any other networks later trained on the same task.