Science and engineering
Machine learning for quantum computing
Quantum machine learning trains parameterised quantum circuits as classifiers and kernels, while classical machine learning helps quantum computing by calibrating devices, decoding errors and characterising quantum states. Fair benchmarks compare every quantum model with tuned classical baselines.
How machine learning is used in quantum computing
Decoding quantum errors
Error correction spreads one logical qubit over many physical qubits and repeatedly measures checks. A decoder must turn these noisy measurements into the right correction quickly, and learned decoders can adapt to the real noise of a device.
Real exampleAlphaQubit (Bausch et al., Nature, 2024)Developed a recurrent, transformer-based neural network that decodes the surface code and outperformed other state-of-the-art decoders on real data from Google's Sycamore quantum processor.
Quantum kernels and variational classifiers
Quantum machine learning maps data into quantum states and trains parameterised circuits or kernels for classification. It is an active research area, with small experiments on current hardware.
Real exampleSupervised learning with quantum-enhanced feature spaces (Havlíček et al., 2018)Proposed and experimentally implemented a variational quantum classifier and a quantum kernel estimator on a superconducting processor.
Characterising quantum states
Full quantum state tomography needs resources that grow exponentially with system size. Neural networks can learn compact representations of a state from simple measurements.
Real exampleNeural-network quantum state tomography (Torlai et al., 2018)Showed that machine learning can reconstruct highly entangled many-body states, and quantities such as entanglement entropy, from simple experimentally accessible measurements.
Benchmarking quantum models fairly
Claims about quantum models are only meaningful against well-tuned classical baselines on the same data. Systematic benchmarks show where quantum models stand today.
Real exampleBetter than classical? (Bowles, Ahmed and Schuld, 2024)A large benchmark study in classical simulation that found out-of-the-box classical machine learning models overall outperformed the quantum classifiers tested.
Can you learn quantum machine learning?
Yes, but it needs two foundations: linear algebra and quantum computing on one side, and statistics and machine learning practice on the other. Learn Python and linear algebra, classical machine learning, then quantum circuits with Qiskit or PennyLane, and always benchmark against classical models.
- PythonWrite and run circuits with Python frameworks such as Qiskit and PennyLane.Learn it in Python
- Linear algebraQubits, gates and measurements are vectors, unitary matrices and projections.Learn it in Maths
- Classical machine learningKernels, support vector machines and neural networks are both the baselines and the building blocks of quantum machine learning.Learn it in Machine Learning
- Quantum computing basicsCircuits, entanglement, noise and error correction from an introductory text and framework tutorials.
- Neural networks for quantum hardwareLearned models for decoding, readout, calibration and state characterisation.Learn it in Deep Learning
Moving into machine learning for quantum computing
If you already work in the field
Quantum physicists and engineers already know the hardware, the noise and the theory. Learn machine learning practice, then apply it to a concrete problem such as readout classification, calibration or decoding with device or public data. Quantum hardware companies and research labs hire for these roles.
If you come from data science or ML
Machine learning practitioners need linear algebra in complex vector spaces and the basics of quantum mechanics and circuits. Qiskit and PennyLane tutorials make this concrete, and reading benchmark studies helps avoid overclaiming quantum advantage.
Roles that do this workQuantum software engineerQuantum machine learning researcherQuantum research scientistQuantum error correction researcher
Tools you will use
| Tool | What you use it for |
|---|---|
| PennyLane | Build and train differentiable quantum circuits with familiar ML libraries. |
| Qiskit | Build circuits and run them on simulators or IBM quantum hardware. |
| Qiskit Machine Learning | Use quantum kernels and quantum neural networks with a scikit-learn style interface. |
| Cirq | Design and simulate circuits for Google quantum processors. |
| scikit-learn | Train the classical baselines every quantum model must be compared against. |
Public data to practise on
- PennyLane downscaled MNIST benchmark: MNIST digits reduced to a few features for fair quantum and classical classifier comparisons.
Your first project
Download the downscaled MNIST benchmark from PennyLane Datasets, train a variational quantum classifier on a simulator, and compare it with logistic regression and an RBF support vector machine tuned with the same budget using test accuracy.
Common questions
What is quantum machine learning?
It covers two things: running machine learning models on quantum computers, such as quantum kernels and variational circuits, and using classical machine learning to help build and operate quantum computers. The second is already in practical use.
Does quantum machine learning beat classical machine learning?
Not on practical tasks today. Careful benchmarks find tuned classical models usually match or beat current quantum models, and proven advantages exist only for specially constructed problems.
Do I need to know quantum physics to learn quantum machine learning?
You need the core of quantum computing: qubits, gates, measurement and noise. That is mostly linear algebra and can be learned without a full physics degree.
Which tools are used for quantum machine learning?
PennyLane, Qiskit and Cirq are the main open-source frameworks, often combined with PyTorch or JAX. All allow simulation on a laptop before running on hardware.
Courses for these skills
- Maths for Machine LearningUnderstand derivatives and matrix shapes.26 published lessons · all planned lessons open
- Deep Learning with PyTorchConnect gradients to neural training objectives.Syllabus available · lessons open as produced
- Statistics and Probability for Machine LearningAssess sampling uncertainty.41 published lessons · all planned lessons open
New to programming or maths? Start with Python, or follow the machine learning path in order. Published lessons are free to read; compare Free and Pro.