Quantum ML Classifies with Hyperellipsoids, Not Hyperplanes

This research provides a comparative analysis of the inherent interpretability of standard linear models and single-qubit mixed-state models for binary classification. The core insight is that a single-qubit mixed-state model effectively performs an "ellipsoid version" of linear classification. Instead of defining a hyperplane to separate data points, this quantum-inspired model learns a hyperellipsoid. The paper delves into the consequences of these distinct geometric inductive biases and how each model incorporates different feature importance biases. This characterization aims to offer an accessible entry point into quantum machine learning concepts for individuals already familiar with classical linear classification, making it a valuable pedagogical tool for introducing quantum ML in undergraduate settings.