Introduction
As the semiconductor industry approaches the physical limits of classical scaling—often referred to as the "post-Moore era"—the computing landscape is diversifying. High-performance computing (HPC) increasingly relies on specialized accelerators, including GPUs, TPUs, and emerging quantum processors (QPUs). However, a significant gap exists in quantum computing education. Most existing curricula are rooted in theoretical physics, focusing on subatomic mechanics rather than the systems-level integration required by computer scientists.
This paper presents the design and evaluation of a specialized Master’s level course at TU Wien. The curriculum shifts the focus from "quantum as a standalone machine" to "quantum as a co-processor" within a distributed, hybrid classical-quantum system.
Pedagogical Approach and Learning Outcomes
The course was designed around four primary Learning Outcomes (LOs) based on Bloom’s taxonomy, specifically tailored for students with a background in computer architecture, algorithms, and software engineering rather than advanced physics.
- LO1 (Conceptual Contrast): Distinguish between classical and quantum computational models.
- LO2 (Theoretical Analysis): Identify potential speedups and theoretical improvements offered by hybrid algorithms.
- LO3 (System Decomposition): Analyze the life cycle of hybrid applications and decompose classical problems into quantum-compatible components.
- LO4 (Implementation): Implement and execute quantum algorithms using modern software frameworks.
By utilizing the gate-based model of quantum computing—which mirrors classical logic gates—the curriculum allows students to grasp quantum logic without requiring deep dives into complex Hilbert spaces or wave-particle duality.
Curriculum Design: The Hybrid Focus
The core of the curriculum is the Hybrid Quantum-Classical Execution Model. Unlike traditional quantum algorithms (like Shor’s) that require fault-tolerant hardware, the course emphasizes Variational Quantum Algorithms (VQAs), such as the Variational Quantum Eigensolver (VQE) and the Quantum Approximate Optimization Algorithm (QAOA). These are particularly relevant for the current Noisy Intermediate-Scale Quantum (NISQ) era.
Execution Hierarchies
Students are taught to navigate three distinct execution models:
- Single Execution: A one-off run of a quantum circuit.
- Job Execution: Repeated runs to gather statistical distributions of results.
- Hybrid Execution: A tight feedback loop where a classical optimizer iteratively updates parameters for a quantum circuit.
Tooling and Workflow
Practical implementation is centered on the IBM Qiskit framework and Python. Students interact with real quantum hardware via the cloud and utilize local simulators (including the "RasQberry" device). To model large-scale scientific pipelines, the course introduces the Pegasus workflow management system, treating the QPU as one node in a broader computational graph.
Evaluation and Results
The course was evaluated through three major assignments involving 19 students. These tasks utilized Scikit-learn datasets for Quantum Support Vector Machine (QSVM) classification and custom instances of the Bin-Packing Problem (BPP). For the latter, students had to transform classical optimization constraints into Quadratic Unconstrained Binary Optimization (QUBO) forms suitable for quantum solvers.
The success rates across the Learning Outcomes were as follows:
- LO1 (Physics vs. Classical): 90.1%
- LO2 (Theoretical Improvements): 88.2%
- LO3 (Hybrid Decomposition): 85.8%
- LO4 (Implementation): 88.3%
The data indicates that while students are highly capable of implementing algorithms (LO4), they struggle most with decomposition (LO3). Determining when and why a specific part of a program should be sent to a quantum processor remains the most significant conceptual hurdle for computer science students.
Discussion and Future Work
The results demonstrate that a systems-oriented approach to quantum computing is both viable and effective for computer scientists. However, the authors noted several areas for refinement. The first iteration relied heavily on project-based assignments; future versions will incorporate more rigorous exams to ensure a uniform understanding of theoretical topics like error correction and photonic qubits.
Furthermore, there is a delicate balance between abstracting away physics and providing enough depth. Future iterations may re-introduce concepts like quantum teleportation and computational universality to provide a more comprehensive theoretical foundation. The ultimate goal is to evolve this curriculum into a dedicated "HPC-Quantum" degree program, positioning quantum computing as a standard component of the modern computer scientist's toolkit.
Summary updated