SAMOP 2023 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

QI: Fachverband Quanteninformation

QI 6: Poster I (joint session QI/Q)

QI 6.8: Poster

Monday, March 6, 2023, 16:30–19:00, Empore Lichthof

Leveraging noisy physical observables with machine learning. — •Adisorn Panasawatwong, Ulf Saalman, and Jan-Michael Rost — Max-Planck-Institute for the Physics of Complex Systems

A noisy light pulse containing many frequencies leads to deterministic electron dynamics in the illuminated target, whose response will also look noisy. At first glance, it cannot be distinguished from a random signal which results from fully chaotic dynamics. While the latter contains little information, the former contains valuable information about the target system, even more than its (linear) response to a Fourier-limited single-frequency pulse.

We are developing a machine learning-based approach which can distinguish the two kinds of noisy signals according to their actual information content: their complexity. Without using entropy, we show emergence of information by interpreting the result from auto-encoder.

Knowing the degree of complexity in the signal enables us to develop networks tailored to extract the amount of information about the target which is contained in the noisy observable due to its complexity.