Decentralized Optimization for Multi-Agent Networks Qing Ling - - PowerPoint PPT Presentation

▶

Dec 23, 2023 41 likes •314 views

Decentralized Optimization for Multi-Agent Networks Qing Ling Department of Automation, University of Science and Technology of China (USTC) Joint work with Wotao Yin (UCLA), Wei Shi and Kun Yuan (USTC) 2014 Workshop on Optimization for Modern

SLIDE 1

Decentralized Optimization for Multi-Agent Networks

1 Qing Ling Department of Automation, University of Science and Technology of China (USTC) Joint work with Wotao Yin (UCLA), Wei Shi and Kun Yuan (USTC) 2014 Workshop on Optimization for Modern Computation 2014/09/02

SLIDE 2

2

Outline

SLIDE 3

3

Multi-agent networks

SLIDE 4

4

Decentralized consensus optimization

SLIDE 5

5

Example: target localization

SLIDE 6

6

Decentralized versus distributed optimization

SLIDE 7

7

Related work

SLIDE 8

8

Assumptions

SLIDE 9

9

Decentralized gradient descent (DGD)

SLIDE 10

10

Mixing matrix

SLIDE 11

11

Existing convergence analysis

SLIDE 12

Can we reach consensus?

12

SLIDE 13

Essence of DGD

13

SLIDE 14

When gradients are bounded?

14

SLIDE 15

15

Where to converge and how fast?

SLIDE 16

16

Concluding DGD

SLIDE 17

17

EXact firsT-ordeR Algorithm (EXTRA)

SLIDE 18

18

Mixing matrices

SLIDE 19

19

Limit properties

SLIDE 20

20

Explanations of EXTRA

SLIDE 21

21

Sublinear convergence

SLIDE 22

Linear convergence

22

SLIDE 23

23

Simulation settings

SLIDE 24

24

Simulation of DGD and EXTRA

SLIDE 25

25

Concluding EXTRA

SLIDE 26

26

Future research directions

SLIDE 27