Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
QI: Fachverband Quanteninformation
QI 14: Quantum Information Poster Session
QI 14.19: Poster
Wednesday, March 11, 2026, 18:00–21:00, P4
Hybrid Quantum-Classical Walks with Applications to Machine Learning — •Adrián Marín Boyero1, Daniel Manzano Diosdado1, and Carlos Cano Gutiérrez2 — 1Department of Electromagnetism and Condensed Matter Physics, Avenida de la Fuente Nueva, 18071 Granada, University of Granada, Granada, Spain — 2Institute Carlos I for Theoretical and Computational Physics, Avenida de la Fuente Nueva, 18071 Granada, University of Granada, Granada, Spain
Complex systems can often be described as networks, where nodes represent elements and edges represent interactions. Classical Random Walks (CRWs) are widely used for machine learning tasks on such structures. In this work, we introduce a Hybrid Quantum-Classical Walk (HQCW) model based on a modified Lindblad equation. This framework interpolates between quantum coherence and classical jumps through a parameter α ∈ [0,1], generating node-visit trajectories whose statistics recover the Lindblad dynamics. The trajectories are generated with a Monte Carlo method known as Quantum Jumps, which reproduces the statistics of the underlying open system evolution. We use these trajectories for Graph Representation Learning and show that HQCWs break node ranking degeneracies, correctly recover communities, and separate node clusters in synthetic networks, even at low embedding dimension d. Performance peaks in a near-classical regime (α ≈ 0.8), capturing the global graph structure effectively. Our results highlight HQCWs as a scalable quantum-inspired tool for network analysis and machine learning.
Keywords: Lindblad Master Equation; Quantum-Inspired Algorithm; Graph Representation Learning; Node Embeddings; Complex Networks