QML Workspace (GSoC Tasks)

This repository contains multiple quantum/ML tasks implemented as part of a GSoC-style project workflow.

Repository structure

• docs/ — project context and research planning notes
• qmaml_hep/ — portable Q-MAML core for HEP-inspired few-shot tasks
• scripts/ — experiment and benchmark runners for Q-MAML
• task1/ — quantum circuit fundamentals (state prep, SWAP test)
• task2/ — classical GNN jet classification (DGCNN vs GAT)
• task3/ — open-ended commentary task (quantum computing/QML perspective)
• task6/ — quantum representation learning with contrastive fidelity loss on MNIST
• test/ — smoke tests for Q-MAML components
• report/ — consolidated LaTeX report for the full workspace

Task highlights

Task 1: Quantum Computing Part

• Implemented 5-qubit circuit with Hadamard/CNOT/SWAP/RX operations
• Implemented SWAP-test circuit to estimate overlap between two 2-qubit states
• Generated diagrams and outputs in task1/

Task 2: Classical GNN for Quark/Gluon Classification

• Graph construction from jet point clouds
• Two architectures implemented:
  • DGCNN (task2/src/dgcnn.py)
  • GAT (task2/src/gat_net.py)
• Training/evaluation pipeline in task2/src/train.py
• Figures in task2/figures/, checkpoints in task2/models/, logs in task2/logs/

Task 3: Open Task

• Personal technical commentary on quantum computing/QML
• Includes algorithm/software perspective and proposed directions

Task 6: Quantum Representation Learning

• MNIST pair sampling with same/different class labels
• Trainable image-to-quantum-state embedding
• SWAP-test fidelity circuit
• Contrastive objective to push same-class fidelity up and different-class fidelity down
• Outputs in task6/outputs/ and task6/outputs_strong/

Q-MAML Extension (Implemented Here)

• Added reusable package in qmaml_hep/ with:
  • synthetic task generation (data.py)
  • variational quantum classifier (model.py)
  • MAML-style inner/outer optimization (qmaml.py)
  • classical + quantum baselines (baselines.py)
• Added execution scripts in scripts/:
  • python -m scripts.run_experiment
  • python -m scripts.run_benchmarks
  • python -m scripts.summarize_benchmarks
• Added smoke test: python -m unittest test/test_qmaml_smoke.py -v
• Setup dependencies via: pip install -r requirements-qmaml.txt

Consolidated report

A full LaTeX report is provided at:

• report/results_report.tex
• compiled PDF: report/results_report.pdf

Build (PowerShell):

cd report
.\build_report.ps1

If script execution is restricted, run directly:

pdflatex -interaction=nonstopmode -output-directory report report\results_report.tex
pdflatex -interaction=nonstopmode -output-directory report report\results_report.tex

Repository governance

• License: LICENSE (MIT)
• Contribution workflow: CONTRIBUTING.md