Beyond the course

The frontier worth knowing about on both sides of the classical-quantum divide, ordered roughly by relevance to radio signal intelligence.

Classical machine learning

Transformers, self-supervised learning, and live deployment used to live on this page - they are now the full Phase 7. What remains here is the frontier beyond it.

Also worth a look

• Reinforcement learning - the ML behind cognitive radio and dynamic spectrum access.
• Generative models (diffusion, VAEs) - used for RF data augmentation and channel modeling.
• Uncertainty and out-of-distribution detection - a signal classifier that knows when it is looking at something new is far more useful than one that guesses confidently.

Quantum

Quantum RF sensing was the sleeper hit of this page - it is now the full Phase 8. What remains here is the longer-horizon quantum landscape.

Quantum error correction

The bridge from NISQ to fault tolerance, and the field's biggest open engineering problem. Surface codes and the decoder problem - where, incidentally, machine learning is used to decode quantum errors. Nielsen & Chuang ch. 10 is the classic entry.

Tensor networks

The mathematics of efficiently representing quantum states classically - the tool used both to simulate quantum circuits and, increasingly, as ML models in their own right (matrix product states as classifiers). The deep answer to "when is quantum hard to fake classically?"

Also worth a look

• QAOA and variational optimization - the other big NISQ algorithm family; relevant to spectrum allocation problems.
• Error mitigation - zero-noise extrapolation and friends; the practical craft of getting signal out of noisy hardware, essential for any near-term QML result.
• Post-quantum cryptography - not quantum computing, but the reason your RF and security worlds will collide with it; know what Shor's algorithm actually threatens.
• Quantum signal processing (the algorithm family) - the term from your original list also names a technique behind modern quantum algorithms (QSP/QSVT); deep, beautiful, and a good third-year topic.

Staying current

• arXiv: skim quant-ph and eess.SP weekly; read one paper properly per week - the three-pass method from Week 11.
• PennyLane demos and blog - new QML techniques usually land there first with runnable code.
• IBM Quantum and Qiskit blogs - hardware roadmap reality checks.
• GNU Radio Conference (GRCon) talks - free on YouTube; where ML-for-RF practitioners actually publish.
• Benchmark tracking - Papers With Code shut down in 2025; for current RadioML state of the art, search arXiv and Hugging Face Papers, or the archived leaderboard data at github.com/paperswithcode/paperswithcode-data.