Disjunctive Threshold Networks for Tabular Data Classification

While neural networks have been achieving increasingly significant excitement in solving classification tasks such as natural language processing, their lack of interpretability becomes a great challenge for to be deployed certain high-stakes human-centered applications. To address this issue, we propose new approach generating interpretable predictions by inferring simple three-layer network with threshold activations, so that it can benefit from effective training algorithms and at the same time, produce human-understandable explanations results. In particular, hidden layer neurons proposed model are trained floating point weights binary output activations. The neuron is also trainable logic function implements disjunctive operation, forming logical-OR first-level functions. This using state-of-the-art methods achieve high prediction accuracy. An important feature architecture only greedy algorithm required provide an explanation human-understandable. comparison other explainable decision models, our achieves more accurate on broad set tabular data datasets.