- Generative Models · Diffusion / GANs
- Bayes-Optimal Learning - theory, bounds, & applications
- Computer Vision - 3D scene understanding, high-fidelity synthesis
- Optimization for Deep Networks - MPS-friendly PyTorch pipelines
| Repo | What it does | Key tech |
|---|---|---|
bolt-loss |
Reference code for “Bayes Optimal Learning Threshold (BOLT) Loss” – plug-and-play loss to tighten Bayes error bounds in classifiers & GANs. | PyTorch Lightning • Hydra • W&B |
quantum-lstm |
Hybrid Quantum-Classical LSTM built with Torch-Quantum | Torch-Quantum • Qiskit |
quantum-transformer |
Hybrid Quantum-Classical Transformer implemented with Torch-Quantum | Torch-Quantum • Qiskit |
Python • PyTorch • Lightning • Hydra • W&B
CUDA / MPS • NumPy • Matplotlib • LaTeX • Git • Docker
- ICASSP 2025 — Universal Training of Neural Networks to Achieve Bayes-Optimal Classification
- Built a neural adaptive streaming framework (“Pensieve”) clone with -8 % rebuffering vs. baseline.
- Reviewer for IJCNN 2025 (Deep Generative Models track).
- Email: tavasoli@umich.edu
- LinkedIn: linkedin.com/in/mohammadreza-tavasoli-naeini-88baa992
- Google Scholar: scholar profile
Built with ❤️ and ☕ | Last updated: 2025‑05‑15