LIGO Document T2200196-v1

Developing Deep Learning Solutions for Lock Acquisition

Document #:
LIGO-T2200196-v1
Document type:
T - Technical notes
Other Versions:
Abstract:
This project will look into investigating and developing deep learning techniques to approach the problem of LIGO’s lock acquisition. Specifically we will look into leveraging modern techniques in attention based learning to help estimate the state of the mirrors given optical signals from the Power Recycled Michelson configuration (PRMI). We will also look into the usage of deep reinforcement techniques and how one might craft a machine learning model that is agnostic to any kind of setup for various degrees of freedom. In exploring these techniques, should any approach prove successful, the impact would directly help improve LIGO’s total operational time with an upper bound of improvement of 12%, which will help accelerate the rate at which gravitational wave events are detected
Files in Document:
Other Files:
Author Groups:
Keywords:
SURF22
Notes and Changes:
Please see gitlab repository for the details and demonstrations. https://gitlab.com/gabrielevajente/prmi-ml/-/tree/main/

DCC Version 3.4.3, contact Document Database Administrators