DPG Phi
Verhandlungen
Verhandlungen
DPG

Berlin 2024 – scientific programme

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

QI: Fachverband Quanteninformation

QI 9: Quantum Machine Learning and Classical Simulability

QI 9.2: Talk

Tuesday, March 19, 2024, 10:00–10:15, HFT-FT 101

Can a neural network fake a Boson Sampler? — •Martina Jung1, Martin Gärttner1, and Moritz Reh1, 21Friedrich-Schiller-Universität, Jena, Deutschland — 2Universität Heidelberg, Heidelberg, Deutschland

Originally defined to demonstrate quantum supremacy, Boson Sampling and its simulation have become an own field of research. The simulation of a Boson Sampler is an - per-construction - classically intractable problem due to the computational complexity of its distribution. A statistics-based approach to learn the probability distribution of a sampling process faces issues like sparse data and highly correlated output configurations. These problems are reminiscent of natural language processing (NLP) tasks where a neural network is trained to respond to a query. Indeed, NLP models like a recurrent neural network (RNN) decompose the task by learning to sequentially predict the next word based on the preceding sequence of words. Transferring this concept to the bosonic Fock space, we train a RNN to simulate a Boson Sampler by predicting the conditional probabilities related to input-output configurations. The model's ability to extrapolate is tested on input sequences of lengths beyond the ones seen during the training.

Keywords: Neural Network; Boson Sampling; Machine Learning

100% | Mobile Layout | Deutsche Version | Contact/Imprint/Privacy
DPG-Physik > DPG-Verhandlungen > 2024 > Berlin