INQA Conference 2025: Sebastian Schulz - Forschungszentrum Jülich

Value of the Information & Strength of the Argument

The talk provides a clear and well-structured argument for why quantum annealing fails and how LDA overcomes these limitations. The value lies in the novel approach of modifying the problem Hamiltonian based on learned features, which is a practical alternative to changing the driver or schedule. The argumentation is supported by detailed analysis of the energy spectrum and landscape, and by benchmarking results on real hardware. The speaker effectively explains complex concepts and justifies each design choice, making a strong case for the method’s efficacy.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high, with a clear methodology and benchmarking against multiple classical and quantum solvers. The sources are primarily the speaker’s own research and the D-Wave system, with no external citations provided in the description. The title accurately reflects the content, which is a conference presentation. The lack of peer-reviewed references and the limited number of problem instances tested slightly reduce the rigor score.

The title accurately reflects the content, which is a conference presentation by Sebastian Schulz on quantum optimization research.

8/10

The talk presents original research with detailed technical explanations, benchmarking on real quantum hardware, and references to classical solvers. The methodology is clearly described, but the lack of peer-reviewed publication details and the limited sample size of demonstrations slightly reduce the score.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and thesis: quantum annealing alone often fails on hard optimization problems.
• 02:00Overview of quantum annealing and the three tuning knobs: schedule, driver, and problem Hamiltonian.
• 05:00Examples of failures: spin glass, factoring, and aircraft scheduling problems.
• 10:00Introduction to iterative quantum annealing and why schedule and driver modifications are limited.
• 15:00Explanation of Learning-Driven Annealing (LDA) and its feature-based Hamiltonian modification.
• 20:00Detailed analysis of energy spectrum and avoided level crossings, relating them to energy landscape features.
• 25:00Demonstration of LDA on spin glass and factoring problems, showing convergence to solutions.
• 30:00Benchmarking results against classical solvers, showing competitive performance.
• 35:00Discussion of global search protocol and how it manipulates many-body localization transitions.
• 40:00Conclusion and outlook for practical quantum optimization.

• D-Wave Advantage system — The quantum annealing hardware used for the experiments.

• Quantum annealing for combinatorial optimization — General review of quantum annealing and its applications.

• Reverse annealing and thermal relaxation — The speaker mentions that reverse annealing often works through thermal relaxation, which is a limitation.