Interpretation of Prediction Models Using the Input Gradient

TL;DR

This paper proposes a straightforward method for interpreting complex machine learning models by analyzing the input gradients, demonstrated on neural networks in natural language processing.

Contribution

It introduces a simple, model-agnostic approach to interpret predictions using input derivatives, applicable to both regression and classification tasks.

Findings

Effective interpretation of neural networks via input gradients

Applicable to various models including convolutional and multi-layer networks

Enhances understanding of model behavior in NLP tasks

Abstract

State of the art machine learning algorithms are highly optimized to provide the optimal prediction possible, naturally resulting in complex models. While these models often outperform simpler more interpretable models by order of magnitudes, in terms of understanding the way the model functions, we are often facing a "black box". In this paper we suggest a simple method to interpret the behavior of any predictive model, both for regression and classification. Given a particular model, the information required to interpret it can be obtained by studying the partial derivatives of the model with respect to the input. We exemplify this insight by interpreting convolutional and multi-layer neural networks in the field of natural language processing.