QubeML

Status: frozen Category: research Owner: alawein Visibility: public Purpose: Quantum machine learning experiments and research. Next action: continue

About

Educational notebooks for quantum computing and materials informatics. Six tool modules covering Qiskit, Cirq, and PennyLane for quantum algorithms, plus PyTorch, scikit-learn, and Kwant for materials modeling.

Public value

QubeML is a research-teaching portfolio candidate: it combines quantum computing tutorials, materials informatics, and notebook-first explanation in a way that is legible to graduate students and technical reviewers. Public polish should focus on reproducible notebooks, dataset provenance, dependency versions, and clear separation between educational examples and original claims.

Features

• Quantum Computing -- VQE for molecular ground states, custom gates, noise simulation, quantum kernels
• Materials Informatics -- Crystal graph neural networks, PCA on materials datasets, 2D material transport
• Tutorial Notebooks -- Jupyter notebooks designed for graduate students and researchers
• Google Colab Support -- All notebooks work in Colab's free tier

Modules

Module	Key Implementations
Qiskit	VQE ground states, ansatz comparison, basis set effects
PyTorch	CGCNN for band gaps, descriptor engineering
Scikit-learn	Materials Project queries, feature importance
Kwant	Graphene ribbons, MoS2 transistors
Cirq	Error mitigation, qubit calibration
PennyLane	Quantum embeddings, kernel methods

Installation

git clone https://github.com/alawein/qubeml.git
cd qubeml
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

Usage

Quantum Chemistry (quantum_computing/qiskit/):

• Build H2 molecule, run VQE with UCCSD ansatz
• Compare to exact diagonalization
• Basis set convergence study

Graph Neural Networks (materials_informatics/pytorch/):

• Load crystal structures from CIF
• Build graph representation
• Train CGCNN on Materials Project data

Transport (materials_informatics/kwant/):

• Graphene nanoribbon conductance
• MoS2 field-effect transistor
• Strain effects on band structure

Project Structure

qubeml/
├── quantum_computing/
│   ├── qiskit/        # VQE tutorials, molecule examples
│   ├── cirq/          # Gate decomposition, error models
│   └── pennylane/     # Quantum ML demos
├── materials_informatics/
│   ├── pytorch/       # GNN implementations
│   ├── scikit_learn/  # Classical ML pipelines
│   └── kwant/         # Transport simulations
├── src/               # Utilities (descriptors, plotting)
└── tests/             # Unit tests

Testing

python -m pytest tests/ -v

Data boundaries

Materials Project queries and generated notebook outputs should document their source, access requirements, and regeneration path. Keep API keys, private datasets, large generated artifacts, and machine-local notebook outputs out of committed examples unless they are intentionally public and reproducible.

References

• Qiskit Textbook: https://qiskit.org/textbook/
• Materials Project: https://materialsproject.org/
• CGCNN paper: Xie & Grossman, Phys. Rev. Lett. 120, 145301 (2018)

License

MIT License -- see LICENSE.

Ownership

• Maintainer: @alawein
• Support: GitHub Issues