Limitations of Classical Machine Learning
This project is part of the introductory module of a Quantum Machine Learning course. It is designed to help students explore the boundaries of Classical Machine Learning (ML) when applied to quantum-inspired data.
Students implement three classical models — Perceptron, k-Nearest Neighbors (k-NN), and Support Vector Machine (SVM) — and train them on a synthetic dataset that simulates qubit states on the Bloch sphere.
Each model is evaluated under four conditions:
Access to the full dataset (ideal case)
Access to only partial features (simulating limited quantum measurement)
Training with added Gaussian noise (simulating decoherence)
A single training epoch (simulating the quantum no-cloning theorem)
Since this project occurs before specific QML algorithms are introduced, students are encouraged to reflect on how Quantum Machine Learning might address these limitations, without direct QML baselines for comparison.
