Quantum Machine Learning
Biamonte et al. (Nature, 2017)
The review that defined the QML research programme. Read it critically and pair it with the dequantization literature, which has since removed several of its headline speedups.
Advanced · 5 resources
Quantum machine learning
Using quantum circuits as trainable models or kernels — plus the sober literature on data loading bottlenecks and classical "dequantization" results that erase many claimed speedups.
Learn first:Variational algorithms (VQE & QAOA)
Why it matters
QML is the most overhyped corner of an already hyped field. It is worth learning precisely so you can tell the small number of interesting results from the noise.
After this you will be able to
- Build quantum kernel and variational classifier models
- Explain the data-loading bottleneck
- Explain what dequantization results imply for QML claims
Start here
3 best places to start
Hand-picked and ordered. If you only have time for one, take the first.
PennyLane demonstrations
Xanadu
Several hundred runnable, paper-linked notebooks on variational algorithms, quantum chemistry and QML. The closest thing the field has to a "reproduce this result" library.
TensorFlow Quantum
Google
Hybrid quantum-classical models built as Keras layers, for people who already know TensorFlow. The tutorials are the fastest route from ML experience into QML.
Also covering this
2 more resources
Quantum algorithm for solving linear systems of equations
Harrow, Hassidim & Lloyd (2008)
HHL, the source of most "exponential speedup for machine learning" claims. The caveats — state preparation, condition number, readout — matter more than the theorem, and are all in the paper.
Variational Quantum Algorithms
Cerezo et al. (2020)
The definitive review of the NISQ-era workhorse: ansatz design, optimisers, barren plateaus and applications, with the open problems stated clearly.