PonderNet: Learning to Ponder
This addresses the problem of inefficient computation in neural networks for AI researchers, offering a novel approach to adaptive computation.
The paper tackled the limitation of standard neural networks where computation grows with input size rather than problem complexity, introducing PonderNet to adapt computation based on complexity, resulting in improved performance on synthetic problems, matching state-of-the-art results with less compute on a QA dataset, and achieving state-of-the-art on a reasoning task.
In standard neural networks the amount of computation used grows with the size of the inputs, but not with the complexity of the problem being learnt. To overcome this limitation we introduce PonderNet, a new algorithm that learns to adapt the amount of computation based on the complexity of the problem at hand. PonderNet learns end-to-end the number of computational steps to achieve an effective compromise between training prediction accuracy, computational cost and generalization. On a complex synthetic problem, PonderNet dramatically improves performance over previous adaptive computation methods and additionally succeeds at extrapolation tests where traditional neural networks fail. Also, our method matched the current state of the art results on a real world question and answering dataset, but using less compute. Finally, PonderNet reached state of the art results on a complex task designed to test the reasoning capabilities of neural networks.1