Soshi Mizutani, Machine Learning Based Forecasting and Its Application for Quantum Materials

Date

2020年1月17日 (金) 16:00

Location

Lab3 C209

Description

Please note the room change - the seminar is at C209

Speaker: Soshi Mizutani, Theory of Quantum Matter (Shannon) Unit

Title: Machine Learning Based Forecasting and Its Application for Quantum Materials

Abstract: Forecasting is used in many different area, ranging from weather forecast to stock market price prediction; it is to predict the future based on the past and present data. In natural sciences, time-dependent data are everywhere, such as animal movements and neuron firings, and we want to ‘forecast’, that is, know what happens next. Human scientists are to find the theory or model governing these events and understand their principles. But can machine learn to do this in their own way?

In this talk, I will first present the fundamentals of forecasting and machine learning, with some widely used models and algorithms. Then, I will show some examples in quantum materials, where time-dependent correlators of quantum spin systems are first generated for small time window by numerical simulation and then later extended by machine learning based forecasting. By applying this method, we are able to obtain predictions for experiment with better energy resolution than ever achieved before.

 

Join us for discussion with free pizza and drinks after the talk! 

All-OIST Category: 

Subscribe to the OIST Calendar: Right-click to download, then open in your calendar application.