Two Instances of Interpretable Neural Network for Universal Approximations
This work addresses the need for interpretable and robust neural networks in machine learning, though it appears incremental as it builds on existing universal approximation concepts.
The paper introduces two interpretable neural network architectures, TNN and SQANN, that achieve universal approximation with resistance to catastrophic forgetting and the ability to detect out-of-distribution samples through interpretable activation fingerprints.
This paper proposes two bottom-up interpretable neural network (NN) constructions for universal approximation, namely Triangularly-constructed NN (TNN) and Semi-Quantized Activation NN (SQANN). Further notable properties are (1) resistance to catastrophic forgetting (2) existence of proof for arbitrarily high accuracies (3) the ability to identify samples that are out-of-distribution through interpretable activation "fingerprints".