Quantum Machine Learning with IBM Quantum (openHPI)

A focused introduction to quantum machine learning (QML) developed by the Hasso Plattner Institute together with IBM Quantum Research. The course is free on the openHPI platform and self-paced. It covers the genuine intersection of quantum computing and machine learning, avoiding hype by grounding every concept in working Qiskit code.

This is one of the few free courses that honestly addresses what quantum machine learning can and cannot do on near-term hardware, while still giving you the tools to implement real QML models yourself.

What you’ll learn

• Introduction to quantum machine learning: what QML is, what it is not, and an honest overview of where quantum advantage in machine learning is (and is not) expected
• Classical machine learning review: supervised learning, loss functions, gradient descent, and support vector machines, at the level needed to understand their quantum counterparts
• Quantum support vector machines (QSVMs): quantum kernel methods, the feature map circuit that defines the kernel, and how the kernel matrix is estimated on hardware
• Variational quantum classifiers (VQC): parameterized quantum circuits as machine learning models, the parameter-shift rule for computing gradients on quantum hardware, and training with classical optimizers
• Quantum generative adversarial networks (QGANs): the GAN framework applied to quantum circuits, the generator and discriminator architecture, and what QGANs can generate
• Quantum Boltzmann machines: the structure of quantum Boltzmann machines, their training procedure, and applications in generative modelling
• Practical Qiskit implementation: the Qiskit Machine Learning library, building feature maps, ansatz circuits, and running QML experiments on simulators and hardware

Course structure

The course is organized into two weeks. Week 1 introduces quantum machine learning, covers support vector machines and their quantum counterpart QSVMs, and implements the variational quantum classifier. Week 2 addresses training quantum models, covers QGANs and quantum Boltzmann machines, and provides practical Qiskit coding exercises.

Video lectures alternate with reading materials and graded assignments. A Record of Achievement is available for learners who complete the assessments.

Who is this for?

• Machine learning practitioners who want to understand what quantum computing brings to their field
• Quantum computing students who already know the basics and want to explore applications in ML
• Researchers at the intersection of quantum computing and data science
• Anyone curious about QML who wants to move beyond blog posts to actual implementation

Prerequisites

Prior knowledge of quantum computing is strongly recommended. The openHPI course “Introduction to Quantum Computing with Qiskit” or IBM Learning’s “Basics of Quantum Information” is good preparation. Basic familiarity with machine learning concepts (supervised learning, classification, loss functions) is helpful. Python and Qiskit experience is needed for the coding exercises.

Hands-on practice

Coding exercises use Qiskit and the Qiskit Machine Learning library:

• Build a quantum feature map circuit and compute the corresponding kernel matrix
• Train a quantum support vector machine on a toy classification dataset
• Implement a variational quantum classifier and train it using gradient descent with the parameter-shift rule
• Build and train a simple QGAN to generate samples from a target distribution
• Compare classification results between classical SVM and QSVM on the same dataset

Why take this course?

IBM Quantum Research produced the core content, which means the course reflects the state of the art in QML research rather than popularised versions. The coverage of QSVMs and quantum kernel methods is particularly valuable: this is the area where rigorous theoretical arguments for quantum advantage in ML have been proposed, and the course explains the theory clearly alongside the implementation.

The self-paced format and free access make it easy to work through on your own schedule. Combined with the Qiskit focus, you finish with working code you can adapt for your own QML experiments.