Sloppy and Genetic Algorithms for Low-Emittance Tuning at CESR Ivan - - PowerPoint PPT Presentation

▶
sloppy and genetic algorithms for low emittance tuning at
SMART_READER_LITE
LIVE PREVIEW

Sloppy and Genetic Algorithms for Low-Emittance Tuning at CESR Ivan - - PowerPoint PPT Presentation

Mar 12, 2024 135 likes •1.11k views

Sloppy and Genetic Algorithms for Low-Emittance Tuning at CESR Ivan Bazarov, William Bergan, Cameron Duncan, Acknowledgements: David Rubin, Jim Sethna DOE DE-SC0013571 Cornell University CESR Sloppy Models Problem Statement Minimize one

slide-1
SLIDE 1

Sloppy and Genetic Algorithms for Low-Emittance Tuning at CESR

Ivan Bazarov, William Bergan, Cameron Duncan, Acknowledgements: David Rubin, Jim Sethna DOE DE-SC0013571 Cornell University

slide-2
SLIDE 2

CESR

slide-3
SLIDE 3

Sloppy Models Problem Statement

Minimize one objective (vertical emittance/beam

size)

Large number of decision variables (independent

magnets)

No reliable auxiliary information (dispersion,

coupling)

Reliable model of machine responses (BMAD

simulation)

slide-4
SLIDE 4

Sloppy Models

  • Phys. Rev. E 68 (2003) 021904.
slide-5
SLIDE 5

Simulated Results

slide-6
SLIDE 6

Experimental Results

RCDS - Nucl. Instr. Meth. 726 (2013) 77.

slide-7
SLIDE 7

Conclusions

Knobs provide some improvements Still far from quantum limit Something missing from models?

slide-8
SLIDE 8

Multi-Objective Genetic Algorithm Problem Statement

What if:

  • Competing criteria of optimal machine

performance

  • In regime where model of machine

responses is unreliable

Needed: a model-agnostic search for optimal

performance trade-offs

slide-9
SLIDE 9

c 1 x c 1 y

xy c2, 1 x 0, 1 y

slide-10
SLIDE 10

c 1 x c 1 y

xy c2, 1 x 0, 1 y

slide-11
SLIDE 11

c 1 x c 1 y x dominates o

xy c2, 1 x 0, 1 y

slide-12
SLIDE 12

c 1 x c 1 y

xy c2, 1 x 0, 1 y

slide-13
SLIDE 13

c 1 x c 1 y neither x nor o dominates

xy c2, 1 x 0, 1 y

slide-14
SLIDE 14

c 1 x c 1 y

xy c2, 1 x 0, 1 y

slide-15
SLIDE 15

c 1 x c 1 y set of non-dominated points

xy c2, 1 x 0, 1 y

slide-16
SLIDE 16

1 Objective A 1 Objective B parent population

How it works

genetic algorithm (spea2) toy example

slide-17
SLIDE 17

1 Objective A 1 Objective B

  • ffspring

genetic algorithm (spea2) toy example

slide-18
SLIDE 18

1 Objective A 1 Objective B survivors (parents of the next generation)

genetic algorithm (spea2) toy example

slide-19
SLIDE 19
slide-20
SLIDE 20
slide-21
SLIDE 21
slide-22
SLIDE 22
slide-23
SLIDE 23
slide-24
SLIDE 24
slide-25
SLIDE 25
slide-26
SLIDE 26
slide-27
SLIDE 27
slide-28
SLIDE 28
slide-29
SLIDE 29
slide-30
SLIDE 30
slide-31
SLIDE 31
slide-32
SLIDE 32
slide-33
SLIDE 33
slide-34
SLIDE 34
slide-35
SLIDE 35
slide-36
SLIDE 36
slide-37
SLIDE 37
slide-38
SLIDE 38
slide-39
SLIDE 39
slide-40
SLIDE 40
slide-41
SLIDE 41
slide-42
SLIDE 42
slide-43
SLIDE 43
slide-44
SLIDE 44
slide-45
SLIDE 45
slide-46
SLIDE 46
slide-47
SLIDE 47
slide-48
SLIDE 48
slide-49
SLIDE 49
slide-50
SLIDE 50
slide-51
SLIDE 51
slide-52
SLIDE 52
slide-53
SLIDE 53
slide-54
SLIDE 54
slide-55
SLIDE 55
slide-56
SLIDE 56
slide-57
SLIDE 57
slide-58
SLIDE 58
slide-59
SLIDE 59
slide-60
SLIDE 60
slide-61
SLIDE 61
slide-62
SLIDE 62
slide-63
SLIDE 63
slide-64
SLIDE 64
slide-65
SLIDE 65
slide-66
SLIDE 66
slide-67
SLIDE 67
slide-68
SLIDE 68
slide-69
SLIDE 69
slide-70
SLIDE 70
slide-71
SLIDE 71
slide-72
SLIDE 72
slide-73
SLIDE 73
slide-74
SLIDE 74
slide-75
SLIDE 75
slide-76
SLIDE 76
slide-77
SLIDE 77
slide-78
SLIDE 78
slide-79
SLIDE 79
slide-80
SLIDE 80
slide-81
SLIDE 81
slide-82
SLIDE 82
slide-83
SLIDE 83
slide-84
SLIDE 84
slide-85
SLIDE 85
slide-86
SLIDE 86
slide-87
SLIDE 87
slide-88
SLIDE 88
slide-89
SLIDE 89
slide-90
SLIDE 90
slide-91
SLIDE 91
slide-92
SLIDE 92
slide-93
SLIDE 93
slide-94
SLIDE 94
  • Needed: a model-agnostic search for optimal

performance trade-offs

  • We tested an elitist genetic algorithm with re-

sampling on bdad simulations of CESR

  • Solution set shows randomness but

converges in statistics

  • Numerical evidence that power-law fit to

solution set is an unbiased estimate of trade-

  • ff front
slide-95
SLIDE 95

Preliminary Results

slide-96
SLIDE 96

Final Thoughts

 Any real-life online optimization metaheuristic is likely to

be a combination of model-cognizant and model- agnostic parts;

 Machine safety needs to “filter” trial solutions preventing

them from adopting forbidden states;

 Noise handling and maximizing throughput are always

key issues;

 CESR is an ideal platform to deploy new kinds of online

  • ptimization strategies, including AI and stochastic

algorithms.