Quantum Machine Learning Workshop
12-13.04.2025

Quantum Machine Learning (QML) is a discipline seeking to take advantage of quantum mechanical processes to induce or enhance machine learning. It combines in novel ways the concepts and algorithms adopted from Quantum Computing and Machine Learning, and is underpinned by the Quantum Mechanics theory and formalism.

This workshop provides an introduction to Quantum Machine Learning using PennyLane and PyTorch, with hands-on exercises and take-home challenges. The workshop includes four practical sessions that cover the QML concepts, models, and techniques. The sessions explore the development of quantum estimators and classifiers, their training with various optimisers, loss and cost functions, as well as model testing and scoring using variety of metrics. It finally, explains how to create hybrid quantum-classical QML models.

Key information

• Workshop dates: 12-13.04.2025 (2 days)

• Organizers: QPoland & Fundacja Quantum AI

• Duration of classes each day: from 10:30 AM  to 2:00 PM (UTC+2)

• Online meeting platform: Zoom

• Communication platform: Discord

• Instructors: Jacob Cybulski, Tomasz Rybotycki, Sebastian Zając 

• Who can attend? The event is only for people who are above 18 years old. It is also mandatory to register and be invited. We also highly recommend checking the "Prerequisite Knowledge: section.

• Registration form (will be open until 11.04.2025, EoD UTC+2): https://forms.gle/PxNqWuzuw3Qy5t9q6 

• Certification: You can get a certificate if you take an online quiz based on the work covered by the workshop activities and complete it successfully within 1 week of the event. 

Anticipated schedule

Note: All challenges have been designed as self-study activities outside the workshop

1. Workshop Day 1 (3.5 hours) / 1 instructor + 1 demonstrator + mentors

• Welcome: Introduction, plan for day 1 and rules of engagement (10 mins)

• Session 1: QML foundation (80 mins)

  • Presentation: QML concepts, models and techniques (30 mins)

  • Demo: PennyLane basics (10 mins)

  • Workshop: Simple models + experiments in model improvement (40 mins)

    • Tasks: Sine function fitting, model refining, data reuploading, local/global cost

    • Challenge: Writing own optimiser and loss functions

• Coffee Break (30 mins)

•  Session 2: QNN estimator development (80 mins)

  • Presentation: Quantum models, gradients, loss / cost and optimisers (30 mins)

  • Demo: QNN estimator in PennyLane (10 mins)

  • Workshop: QNN estimator + experiments in model improvement (40 mins)

    • Tasks: Fitting linear data, PennyLane models, training and test scoring

    • Challenge: Create a model for real-estate house valuations

• Summary: Reflection and day 2 preview (10 mins)

• Optional work to be given: Self-study and challenges

2. Workshop Day 2 (3.5 hours) / 1 instructor + 1 demonstrator + mentors

• Summary: Reflection and plan for day 2 (10 mins)

• Session 3: Intermediate QML (80 mins)

  • Presentation: Quantum classifiers, metrics, model training and testing (30 mins)

  • Demo: PennyLane classifier in PennyLane with PyTorch (10 mins)

  • Workshop: Quantum classifiers + experiments in model improvement (40 mins)

    • Tasks: Automotive risk assessment, data preparation, QNN classifier

    • Challenge: Quantum multinomial classification 

• Coffee Break (30 mins)

• Session 4: Advanced QML (80 mins)

  • Presentation: Hybrid quantum-classical model development (30 mins)

  • Demo: Hybrid model in PennyLane with PyTorch (10 mins)

  • Workshop: Hybrid model + experiments in model improvement (40 mins)

    • Tasks: Redevelop the previous classifier as a hybrid quantum-classical model

    • Challenge: Develop and test a quantum reservoir model 

• Summary: Reflection and congrats (10 mins)

• Optional work to be given: Self-study and challenges

Prerequisite Knowledge

1. Python 3

• Familiarity with: venv or anaconda (virtual environments)

• Intermediate Python with some advance concepts, e.g. classes

• Knowledge of packages: numpy, pandas, scikit-learn, matplotlib, jupyter

• Knowledge of pytorch and gradients would greatly help.

2. Quantum Computing

• Fundamentals: qubits, circuits, gates, quantum state, measurement, Dirac notation

• Maths: complex numbers, matrices / tensors, some statistics and probability theory

• Hands-on: one of the QC platforms (Qiskit, PennyLane, Cirq, ...)

3. Classical Machine Learning

• Models: classification, regression, neural networks

• Model training: optimisers, loss / cost function, gradients

• Model performance: measurements, training vs testing

4. Other knowledge

• Discord - workshop communication platform (notes, networking, discussions, get help and pointers)

• GitHub (demos, exercises and data)

• Pre-reading of presentations for Day 1 and Day 2

Pre-Workshop Preparation (use own computer with recent Linux / Window / macOS)

1.  Download all resources (notes, code and data).

2. Install the recommended Python virtual environment (venv + requirements).

3. Undertake some preliminary exercises and get familiar with:

• PyTorch, tensors, gradients and neural networks: AssemblyAI, “PyTorch Crash Course - Getting Started with Deep Learning”, Jul 2022. https://www.youtube.com/watch?v=OIenNRt2bjg (50 mins)

• PennyLane, functions, circuits, qnodes and measurements: Diego Emilio Serrano, “Basic Introduction to PennyLane”, Feb 2023. https://www.youtube.com/watch?v=MCDHAn-GvA8 (40 mins)

• PennyLane circuit creation and execution for busy people: Isaac De Vlugt, “My first quantum circuit in PennyLane”, Sept 2023. https://www.youtube.com/watch?v=uCm027_jvZ0 (5 mins)

4. Workshop Software Environment (installation instructions and requirements file will be provided):

• Create a virtual environment created with venv or anaconda for Python 3.11 and pip 24.0+

• The following Python packages will be used, for: QML PennyLane 0.40.0 (with PennyLane_Lightning), ML (scikit-learn 1.6.1, pandas 2.2.3, numpy 1.26.4) and torch 2.6.0 (with torchaudio, torchvision, torchsummary, torcheval and torchmetrics), plotting and image processing (matplotlib 3.10.1, plotly 6.0.0, seaborn 0.13.2, pillow 11.1.0), jupyter 1.1.1 (with jupyterlab 4.3.5), and data access (kagglehub 0.3.10, ucimlrepo 0.0.7), plus their dependencies.

Organizers