1. Introduction

T-SHIELD (Transformation-Selective Hidden Input Evaluation for Learning Dynamics) is a regularization technique that approach aims to enhance model interpretability while improving model performance. Specifically, T-SHIELD adds a regularization term to the objective function of a model optimization loss that penalizes the model if it relies too heavily on a small subset of input features.

T-SHIELD is a regularization family that can be used with any model that can be trained using gradient descent. It is implemented in Python and is available as a package that can be installed with this repo.