Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Novel Approach to Explainable AI with Quantized Active Ingredients in Decision Making

A. M. A. S. D. Alagiyawanna, Asoka Karunananda, Thushari Silva, A. Mahasinghe

TL;DR

This paper tackles explainable AI in high-stakes domains by comparing Quantum Boltzmann Machines (QBMs) with Classical Boltzmann Machines (CBMs) on a PCA-reduced MNIST task restricted to digits 0 and 1. It introduces a hybrid quantum-classical QBM with a quantum hidden layer trained alongside a classical RBM baseline, and uses gradient-based saliency for the QBM and SHAP for the CBM to quantify feature attributions. The study reports that QBMs achieve higher accuracy (83.5% vs 54%) and exhibit more focused attributions (lower entropy) in the latent space, suggesting clearer identification of active ingredients behind predictions. The findings indicate that quantum-classical hybrids can improve both predictive performance and interpretability, offering a path toward more trustworthy AI systems and outlining future work to scale to more complex data and integrate hybrid interpretability approaches.

Abstract

Artificial Intelligence (AI) systems have shown good success at classifying. However, the lack of explainability is a true and significant challenge, especially in high-stakes domains, such as health and finance, where understanding is paramount. We propose a new solution to this challenge: an explainable AI framework based on our comparative study with Quantum Boltzmann Machines (QBMs) and Classical Boltzmann Machines (CBMs). We leverage principles of quantum computing within classical machine learning to provide substantive transparency around decision-making. The design involves training both models on a binarised and dimensionally reduced MNIST dataset, where Principal Component Analysis (PCA) is applied for preprocessing. For interpretability, we employ gradient-based saliency maps in QBMs and SHAP (SHapley Additive exPlanations) in CBMs to evaluate feature attributions.QBMs deploy hybrid quantum-classical circuits with strongly entangling layers, allowing for richer latent representations, whereas CBMs serve as a classical baseline that utilises contrastive divergence. Along the way, we found that QBMs outperformed CBMs on classification accuracy (83.5% vs. 54%) and had more concentrated distributions in feature attributions as quantified by entropy (1.27 vs. 1.39). In other words, QBMs not only produced better predictive performance than CBMs, but they also provided clearer identification of "active ingredient" or the most important features behind model predictions. To conclude, our results illustrate that quantum-classical hybrid models can display improvements in both accuracy and interpretability, which leads us toward more trustworthy and explainable AI systems.

A Novel Approach to Explainable AI with Quantized Active Ingredients in Decision Making

TL;DR

This paper tackles explainable AI in high-stakes domains by comparing Quantum Boltzmann Machines (QBMs) with Classical Boltzmann Machines (CBMs) on a PCA-reduced MNIST task restricted to digits 0 and 1. It introduces a hybrid quantum-classical QBM with a quantum hidden layer trained alongside a classical RBM baseline, and uses gradient-based saliency for the QBM and SHAP for the CBM to quantify feature attributions. The study reports that QBMs achieve higher accuracy (83.5% vs 54%) and exhibit more focused attributions (lower entropy) in the latent space, suggesting clearer identification of active ingredients behind predictions. The findings indicate that quantum-classical hybrids can improve both predictive performance and interpretability, offering a path toward more trustworthy AI systems and outlining future work to scale to more complex data and integrate hybrid interpretability approaches.

Abstract

Artificial Intelligence (AI) systems have shown good success at classifying. However, the lack of explainability is a true and significant challenge, especially in high-stakes domains, such as health and finance, where understanding is paramount. We propose a new solution to this challenge: an explainable AI framework based on our comparative study with Quantum Boltzmann Machines (QBMs) and Classical Boltzmann Machines (CBMs). We leverage principles of quantum computing within classical machine learning to provide substantive transparency around decision-making. The design involves training both models on a binarised and dimensionally reduced MNIST dataset, where Principal Component Analysis (PCA) is applied for preprocessing. For interpretability, we employ gradient-based saliency maps in QBMs and SHAP (SHapley Additive exPlanations) in CBMs to evaluate feature attributions.QBMs deploy hybrid quantum-classical circuits with strongly entangling layers, allowing for richer latent representations, whereas CBMs serve as a classical baseline that utilises contrastive divergence. Along the way, we found that QBMs outperformed CBMs on classification accuracy (83.5% vs. 54%) and had more concentrated distributions in feature attributions as quantified by entropy (1.27 vs. 1.39). In other words, QBMs not only produced better predictive performance than CBMs, but they also provided clearer identification of "active ingredient" or the most important features behind model predictions. To conclude, our results illustrate that quantum-classical hybrid models can display improvements in both accuracy and interpretability, which leads us toward more trustworthy and explainable AI systems.
Paper Structure (9 sections, 5 equations, 13 figures)

This paper contains 9 sections, 5 equations, 13 figures.

Table of Contents

  1. Introduction
  2. Methodology
  3. Approach
  4. Reducing Dimensionality
  5. Quantum Hidden Layer Design
  6. Design of Classical Boltzmann Machine
  7. Evaluation of QBM
  8. Results and Discussion
  9. Conclusion

Figures (13)

  • Figure 1: A graphical representation of an example Boltzmann machine. In this example, there are 3 hidden units and 4 visible units patel2020overview.
  • Figure 2: Visualization of a qubit using bloch sphere representation jazaeri2019review.
  • Figure 3: Representation of how shap works NIPS2017_7062.
  • Figure 4: Approach to integrate Quantum Computing with RBM
  • Figure 5: Before and after dimensionality reduction of "0" data from MNIST data set
  • ...and 8 more figures