A Novel XAI-Enhanced Quantum Adversarial Networks for Velocity Dispersion Modeling in MaNGA Galaxies

TL;DR

This work addresses the challenge of predicting MaNGA galaxy velocity dispersion with models that are accurate, robust, and interpretable. It introduces an XAI-enhanced quantum adversarial framework that blends a hybrid quantum neural network with classical layers and an evaluator using LIME explanations, extended by GAN and self-supervised variants. The Vanilla model achieves a balanced set of regression metrics (RMSE $=0.27$, MSE $=0.071$, MAE $=0.21$, $R^2=0.59$) and superior calibration, underscoring the value of hybrid quantum-classical architectures augmented with explainability. The results highlight that while adversarial variations offer selective gains, the Vanilla configuration provides the most reliable and interpretable performance, with implications for scalable quantum-enhanced modeling in astrophysics and beyond.

Abstract

Current quantum machine learning approaches often face challenges balancing predictive accuracy, robustness, and interpretability. To address this, we propose a novel quantum adversarial framework that integrates a hybrid quantum neural network (QNN) with classical deep learning layers, guided by an evaluator model with LIME-based interpretability, and extended through quantum GAN and self-supervised variants. In the proposed model, an adversarial evaluator concurrently guides the QNN by computing feedback loss, thereby optimizing both prediction accuracy and model explainability. Empirical evaluations show that the Vanilla model achieves RMSE = 0.27, MSE = 0.071, MAE = 0.21, and R^2 = 0.59, delivering the most consistent performance across regression metrics compared to adversarial counterparts. These results demonstrate the potential of combining quantum-inspired methods with classical architectures to develop lightweight, high-performance, and interpretable predictive models, advancing the applicability of QML beyond current limitations.

Figures (20)

• Figure 1: Illustration of the distribution of each feature in the Model Building
• Figure 2: PCA (left) and t-SNE (right) plots showing the data structure and clustering in reduced dimensions.
• Figure 3: Feature Importance Plot
• Figure 4: LDA plot of discretized logsigmae classes, showing clear separation between low, medium, and high groups.
• Figure 5: Architecture of the Vanilla QNN-Evaluator model for each Epoch.