eXplainable AI for Quantum Machine Learning

Parametrized Quantum Circuits (PQCs) enable a novel method for machine learning (ML). However, from a computational point of view they present a challenge to existing eXplainable AI (xAI) methods. On the one hand, measurements on quantum circuits introduce probabilistic errors which impact the convergence of these methods. On the other hand, the phase space of a quantum circuit expands exponentially with the number of qubits, complicating efforts to execute xAI methods in polynomial time. In this paper we will discuss the performance of established xAI methods, such as Baseline SHAP and Integrated Gradients. Using the internal mechanics of PQCs we study ways to speed up their computation.

Introduction

Since most ML models are too complex for humans to properly interpret, methods have been devised to study these models and to provide the ability to understand how models arrive at certain predictions. This explanation is valuable for model developers, who need to understand e.g. the limitations from a model on a global perspective to adjust the model architecture or the training data. It is also valuable for model users, who are interested in the factors that drive specific model predictions.

Hence, those methods can improve the overall model quality as well as model acceptance. In addition, governmental initiatives like the European Union's 'Articial Intelligence Act' [1] require that models applied in high-risk areas (e.g. access to education and essential private services) provide a minimum level of transparency for users, which highlights the importance of xAI methods.

In this paper, we will study models based on so-called Parameterized Quantum Circuits (PQCs) [2], which are also known as variational quantum circuits or quantum neural networks [3], and how xAI methods can be adjusted to these type of models.

[…]

Den gesamten Artikel können Sie oben herunterladen.