Quantum Machine Learning Capstone Project Proposal: Design, Implementation, and Evaluation

Table of Contents

1. Project Overview
2. Motivation and Objectives
3. Background and Literature Review
4. Problem Statement
5. Proposed Methodology
6. Dataset Description and Preprocessing
7. Quantum Circuit Design
8. Classical-Quantum Hybrid Integration
9. Model Training and Optimization
10. Performance Evaluation Metrics
11. Hardware and Software Tools
12. Implementation Plan and Milestones
13. Risk Management and Mitigation
14. Ethical and Security Considerations
15. Expected Outcomes
16. Benchmarking and Comparative Study
17. Scalability and Future Extensions
18. Capstone Deliverables
19. Team Roles and Responsibilities
20. Conclusion

1. Project Overview

This capstone project aims to design, implement, and evaluate a quantum machine learning (QML) model for solving a real-world classification or recommendation problem using variational quantum circuits and hybrid quantum-classical learning pipelines.

2. Motivation and Objectives

• Explore the potential of QML in a practical application domain
• Gain hands-on experience with quantum development tools
• Demonstrate viability of hybrid approaches on NISQ devices

3. Background and Literature Review

Survey recent advancements in:

• Variational quantum classifiers (VQC)
• Quantum-enhanced kernels
• Hybrid QML with PennyLane, Qiskit, and TFQ
  Key papers from arXiv, IBM Qiskit Blog, and Nature Quantum Information

4. Problem Statement

Design a quantum machine learning model that performs binary or multiclass classification on a structured or image dataset, and evaluate its accuracy and efficiency against classical baselines.

5. Proposed Methodology

• Preprocess data using classical tools (scikit-learn, pandas)
• Encode features into quantum states
• Construct and train a VQC using parameter-shift gradients
• Benchmark using simulators and QPU execution
• Evaluate robustness, accuracy, and noise tolerance

6. Dataset Description and Preprocessing

• Potential datasets: Iris, Breast Cancer, MNIST (reduced)
• Normalize and reduce to 2–8 dimensions (qubit-friendly)
• Convert labels and encode categorical variables

7. Quantum Circuit Design

• Feature map: angle or amplitude encoding
• Ansatz: TwoLocal, RealAmplitudes, or custom layered entanglement
• Optimizer: COBYLA, SPSA, or gradient descent

8. Classical-Quantum Hybrid Integration

• Use PyTorch, TensorFlow, or JAX for gradient propagation
• Combine quantum layer outputs with classical classifiers
• Train end-to-end with loss minimization

9. Model Training and Optimization

• Apply batching and adaptive learning rates
• Use cross-validation and multiple random seeds
• Log metrics like loss, accuracy, circuit depth

10. Performance Evaluation Metrics

• Accuracy, F1-score, ROC-AUC
• Fidelity of quantum states
• Execution time and shot efficiency

11. Hardware and Software Tools

• PennyLane or Qiskit
• IBM Quantum Experience or Amazon Braket
• Python, NumPy, matplotlib for visualization

12. Implementation Plan and Milestones

• Week 1: Problem finalization, literature review
• Week 2–3: Dataset preparation, circuit design
• Week 4: Model integration, training setup
• Week 5: Simulation testing, tuning
• Week 6–7: Real hardware deployment, analysis
• Week 8: Report writing, poster, and demo

13. Risk Management and Mitigation

• Limited qubit availability → use simulators for tuning
• Hardware queue delays → submit early batches
• Circuit too deep → use compressed ansatz

14. Ethical and Security Considerations

• Respect privacy if using real-world data
• Secure access to cloud quantum providers
• Avoid biased model design via class balancing

15. Expected Outcomes

• Trained QML model with competitive performance
• Comparison against classical ML baseline
• Execution and performance report from real quantum device

16. Benchmarking and Comparative Study

• Compare with SVM, logistic regression, MLP
• Evaluate training time, robustness, and generalization

17. Scalability and Future Extensions

• Extend to image, graph, or time-series data
• Explore quantum GANs or kernel boosting
• Deploy as web app or streamlit dashboard

18. Capstone Deliverables

• Project report
• Python source code
• Quantum circuit visualization
• Presentation and demo script

19. Team Roles and Responsibilities

• Research Lead: Literature review, benchmarking
• Dev Lead: Circuit building and optimization
• Data Analyst: Preprocessing and evaluation
• Report Writer: Documentation and presentation

20. Conclusion

This capstone will provide end-to-end exposure to quantum machine learning from design to deployment. By working with real quantum hardware and simulators, students will build a foundation for future contributions to the quantum AI field.