Machine Learning Meeting Agenda, June 26, 5pm CST
1) Cost function minimization
- Cost function minimization algorithms, what has been done - simulations and experiment (Den)
- NN based feedback algorithm (Masha)
- Adaptive feedback LMS (Den)
2) Finding an optimal cost function
- Creating an optimal cost function for a feedback control of a cavity, in what sense is it optimal? (Jenne)
- Adaptation of a particular cost function to keep RMS of the cavity constant (Den)
- How to convert lock loss minimization, actuator saturations, stability criteria and error minimization to cost function? (All)
- MIMO loop tuning (Rana)
3) Sensor blending
- Blending of GS13 and CPS, L4C and IPC for ISI and HEPI control in optimal way (Den)
Pre-OMC heterodyne and DC-readout (-> p4344, 4376) blending or hierarchical control, is it useful? (Koji)
4) Reinforcement learning algorithms
- Test of CARCA (Den)
- Should we create a library of codes? (All)
5) Lock acquisition states
- Correspondance to seismic motion, experience from i,eLIGO, Virgo, GEO, Tama... (Koji, Rana)
- Seismic radar, k-maps, status of the project (Den)
6) Problems to solve
- experiment - test CARCA online on an oplev (40m)
- experiment - test seismic k-maps (LLO, 40m)
- experiment, simulation - adaptive feedback LMS on OL (40m)
- experiment - DRMI loop tuning (40m)
- simulations - RL algorithm testing
- simulations - NN feedback controller
- simulations - tuning IFO for a particular GW source - adjust control system, squeezing angle, readout type
- simulations - loops crossover tuning - suspensions, frequency stabilization