Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

8
Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui

Transcript of Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Page 1: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Population Based Incremental Learning

Shumeet Baluja

Presented by KC Tsui

Page 2: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Background

• Populations based search, such as GA– Create a probability matrix by counting the number of 1s

and 0s in each gene position– Generate new population using the probability matrix– No information is carried from generation to generation!

• Supervised Competitive learning, e.g. LVQ– Winner-take-all reinforcement learning in ANN– Winner is a kind of prototype of the sample presented

• PBIL = GA + CL– Capture the trend from the best performer

Page 3: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Basic PBIL

P initialize probability vector (each position = 0.5)while (generations++ < limit)

for each vector i dofor each position j do

generate Vi(j) according to P(j)end-doevaluate f(Vi)

end-doVmax = max(f(Vi))update P according to Vmax

if random(0,1] < Pmutate

mutate Pend-if

end-while

Page 4: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Update and Mutation Rules

• Update Rule

• Mutation Rule

– Pmutate = 0.02

= 0.05

*)()0.1(*)()( max jVjPjP ii

*]0.10.0[)0.1(*)()( randjPjP ii

Page 5: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Extensions

• Learning from M top scorer: Adapt the probability matrix– equally according to the M vectors, or

– where there is consensus in, or most, of the M vectors

– according to the rank of the M top vectors

• Learning also from negative sample– move away the worst vector

– modify only those positions where the best and the worst disagree

Page 6: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Some applications

• Function optimization

• Job-shop scheduling

• TSP

• Bin-packing

• Knapsack Problem

• Neural Network weight training

Page 7: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

Some ComparisonPBIL EDO EA

population Fixed size Variable size Fixed size

Probability matrix

One One per parent None

parameters fixed Adaptive Can be adaptive

learning Pos + neg Pos + neg pos

diversity high Very high Gradually decreasing

Search scope Local Local Global & local

Page 8: Population Based Incremental Learning Shumeet Baluja Presented by KC Tsui.

References

• Shumeet Baluja, Population-Based Incremental Learning: a method for integrating genetic search based function optimization and competitive learning, Technical report CMU-CS-94-163. 1994.

• Shumeet Baluja & Rich Caruana, Removing the Genetics from the Standard Genetic Algorithm, ICML’95.

• Shumeet Baluja, An Empirical Comparison of Seven Iterative and Evolutionary Function Optimization Heuristics, Technical Report CMU-Cs-95-193, 1995.