Topics and Goals

Quantum Computing (QC) has gained attention for its promise to substantially accelerate the solution of many computationally intensive tasks. Recommender Systems (RS), which must operate over large-scale, heterogeneous data using complex algorithms, are one of many application domains where QC may offer computational advantages. This tutorial aims to provide an accessible introduction to QC, with a focus on the Quantum Annealing (QA) paradigm. The tutorial is designed for an audience without prior expertise in quantum technologies and presents a practical overview of how RS problems such as community detection can be modeled using the Quadratic Unconstrained Binary Optimization (QUBO) formulation and solved using QA. Participants will be guided through the theoretical foundations and will gain hands-on experience in formulating and running RS-related tasks for quantum annealers

Outline

The tutorial will cover theoretical and practical aspects underneath QC and especially Quantum Annealing (QA), by allowing participants to learn how to code and use quantum annealers to solve optimization problems relevant for RS and IR.

Part 1: QC Foundations (40 min)

The first part offers an accessible introduction to QC, outlining both its potential advantages and limitations. It will also cover the QA paradigm, highlighting key distinctions between classical and quantum approaches to computation and addressing common misconceptions about QC. Topics include:

  • A general overview of QC, including its prospective benefits;
  • A basic introduction to the Quantum Circuit and Quantum Adiabatic computation models;
  • Connections between classical meta-heuristics such as Simulated Annealing and their quantum counterpart, Quantum Annealing, within the QA paradigm;
  • The principle of adiabatic evolution leading a quantum system toward its ground state;
  • The use of the Ising model to represent a quantum system's energy configuration, or Hamiltonian;
  • The structural relationship between the Ising formulation and the Quadratic Unconstrained Binary Optimization (QUBO) model used to solve NP-hard optimization problems.

Part 2: QUBO Formulation (50 min)

This session focuses on translating classical binary optimization tasks into the QUBO formulation. Participants will learn how to express both NP-complete decision problems (such as graph partitioning) and NP-hard optimization problems (such as the quadratic assignment problem) using QUBO models, with attention to how constraints and objective functions are encoded.

Part 3: QC for IR and RS and their Evaluation (30 min)

This part introduces several RS and IR tasks that can be effectively addressed using QA, including Feature Selection, Community Detection, and Clustering. We will examine how to evaluate the effectiveness and efficiency of these quantum algorithms. The session concludes with an overview of the QuantumCLEF lab 2024 and 2025, summarizing key findings and lessons learned.

Part 4: Hands-on (60 min)

This session offers a practical walkthrough on how to interact with quantum annealers, which are accessible via cloud-based platforms. It covers:

  • The underlying architecture and connectivity of quantum annealers;
  • The process of mapping a problem into QUBO form and program it on the quantum hardware through Minor Embedding;
  • The influence of QUBO density on qubit usage and embedding complexity;
  • How to execute quantum programs and retrieve solutions (hands-on);
  • Implementation of Feature Selection and Community Detection algorithms as QUBO problems (hands-on)

Target Audience

The intended audience are researchers and practitioners from RS and IR, as well as from other fields, such as Natural Language Processing, Machine Learning, Big Data, Operations Research, and Optimization. QC and QA are powerful tools that can be applied to tackle problems in many different domains and, while the hands-on part is focused on how to use QA for RS, the problems we consider are quite general and have application to different areas.

This tutorial will be self-contained and has minimal prerequisite knowledge, mainly consisting of being familiar with the basic concepts of RS and IR. For those interested in the hands-on part, basic Python programming skills are required to use the tools provided by D-Wave3 but no specific knowledge of the D-Wave API is required ahead.

Material

Here you can find the slides used during the tutorial:

Here you can find the link to the git repository where you can find the notebooks for the Hands-On part: Link to the repository