Reproduce “Automatic design of quantum feature maps” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Sergio Altares-López, Angela Ribeiro, Juan José García-Ripoll
Researchers from CAR (CSIC-UPM) and IFF-CSIC introduce a method for automatically designing quantum feature maps using a multiobjective genetic algorithm. This approach generates optimized quantum circuits that achieve high classification accuracy while minimizing structural complexity, frequently outperforming classical machine learning algorithms on various datasets.
Key insights
- The genetic algorithm consistently produced highly accurate quantum feature maps that were structurally simple, often unentangled or minimally entangled, even for non-linear classification tasks.
- The specific quantum kernel K(x, x') = Re <Φ(x)|Φ(x')> was found to facilitate sharper separations and easier convergence in numerical simulations compared to its squared magnitude counterpart.
- Simplified, low-entanglement circuits can achieve complex classification through the collective effect of individual qubit contributions, enhancing interpretability by allowing analysis of decision boundaries from sub-circuits.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.