Opened 5 months ago

Last modified 4 months ago

#3107 new enhancement

Implement "Learning ALPS"

Reported by: pfleck Owned by: pfleck
Priority: medium Milestone:
Component: Algorithms.ALPS Version: branch
Keywords: Cc:

Description

Since ALPS is capable of regularly produce new random populations in the bottom layer, we could take information from higher layers and steer the random creation in promising directions.

For instance, in GP, we could vary the symbol and variable frequencies for the random population by analyzing which symbols and variables perform well in the upper layers. Additionally, we could also increase diversity by specifically generate individuals in the search area where the top layers haven't been looking.

Change History (5)

comment:1 Changed 5 months ago by pfleck

r17852 Created branch.

comment:2 Changed 5 months ago by pfleck

r17854 Added a new operator hook for ALPS for adapting the reinitialization strategy.

comment:3 Changed 5 months ago by bburlacu

r17857: Implement crude initial reinitialization concept.

comment:4 Changed 4 months ago by pfleck

r17916 Added missing placeholder for the Reinitialization Strategy.

comment:5 Changed 4 months ago by pfleck

r17947 Added MinimumFrequency and LearningRate parameters to symbol frequency and node impact strategy controllers.

Note: See TracTickets for help on using tickets.