GPPU Seminar

Date

13:00-15:00, January 17th, 2018

Place

Room 303, Science Complex A (H-02) map

Abstract

Investigation of physical well-motivated parameter space plays an important role in new physics (especially supersymmetry) exploration. However, a large-scale scan of high dimensional parameter space under vast experimental constraints is typically a time-consuming and expensive task. In this talk, a new self-learning scan approach, named Machine Learning Scan (MLS), is introduced. This MLS can achieve a fast and reliable exploration of high dimensional parameter space by using machine learning models to evaluate the quality of random parameter sets. As a proof-of-concept, we apply MLS to several benchmark models, including the alignment limit of the minimal supersymmetric model, and find that such a method can significantly reduce the computational cost and ensure the discovery of all survived regions (comparisons with the conventional scan methods are provided). At the beginning of the talk, some basics about Machine Learning is also presented.

Point

GASP 1