Input Switched Affine Networks: An RNN Architecture Designed for Interpretability
This work addresses the need for interpretable models in domains where deployment requires understanding, though it appears incremental in its approach.
The authors tackled the problem of interpretability in neural networks by introducing an RNN architecture without explicit nonlinearities, which achieved reasonable performance on text modeling tasks and allowed greater computational efficiency.
There exist many problem domains where the interpretability of neural network models is essential for deployment. Here we introduce a recurrent architecture composed of input-switched affine transformations - in other words an RNN without any explicit nonlinearities, but with input-dependent recurrent weights. This simple form allows the RNN to be analyzed via straightforward linear methods: we can exactly characterize the linear contribution of each input to the model predictions; we can use a change-of-basis to disentangle input, output, and computational hidden unit subspaces; we can fully reverse-engineer the architecture's solution to a simple task. Despite this ease of interpretation, the input switched affine network achieves reasonable performance on a text modeling tasks, and allows greater computational efficiency than networks with standard nonlinearities.