Deepmind introduces PonderNet, a new algorithm that allows artificial neural networks to learn to think for a while before answering. This improves the ability of these neural networks to generalize outside of their training distribution and answer tough questions with more confidence than ever before.
The time required to solve a problem is not just influenced by the size of inputs but also the complexity. Also, the amount of computation used in standard neural networks is not proportional to the complexity, but rather it’s proportional with size. To address this issue, Deepmind, in its latest research, presents PonderNet, which builds on Adaptive Computation Time (ACT; Graves, 2016) and other adaptive networks.
PonderNet is fully differentiable and can leverage low-variance gradient estimates (unlike REINFORCE). It has unbiased gradient estimates, unlike ACT. It achieves this by reformulating the halting policy as a probabilistic model.
PonderNet is a neural network algorithm that incentivizes exploration over pondering time to improve the accuracy of predictions. Deepmind researchers used PonderNet on their parity task and demonstrated how it can increase computation when extrapolating beyond seen data during training, achieving higher accuracy in complex domains such as question answering and multi-step reasoning.