15 January 2014

Present

Tom, Omri, Bill, Karthik (scribe)

Administrative stuff

Meeting times – Wednesdays 10-12 Starting 22nd January 2014
29th – GECCO deadline.

ULTRA stuff

Bug with ULTRA-padding and push simplification fixed.
Single symbol program issues.
Potentially solve with making all programs sequences.
Single instruction to ULTRA where ULTRA does nothing. #New issue

Push Licensing

Push licensing currently unresolved.
The Eclipse Public License ? The one Clojure uses for now.

GECCO paper stuff

– Differential equations
– Word count
– Push simplification
– (Abstract on calc)
– Finite algebras
Random stuff : Source code addition to GECCO ? Anonymous everything – tricky.
Fischer’s exact test – potential statistical significance test.
Justifications for why some particular statistics were used.
Program self reference interference with simplification.

George

Searching for Quantum algorithms to use GP on. Qgame used? Possibly depends on the need for complex numbers.

Karthik

Lexicase + GSX. Lexicase currently done without push. Possible extension of some GSX stuff to also use Lexicase.

Calc problem updates

Complimentary mate selection.
Reversed lexicase selection.
Errors, Pass/Fail, and Novelty
The Genetics : Recombination for correct functionality.
Tag modifications for combining functionality correctly ?
Small deviations vs Large deviations with ULTRA.
ULTRA causing program size to increase because of the program size average

Clojush simplifcation

Simplifcation at the end of the run
Simplification for reporting
Simplification during run – where simplification is a genetic operator. (Messy courtesy fitness evaluation being built IN to a genetic operator)

Bill

Different differential equation models.
“Non inheritable epigenesis”
Hill climbing and improvement.
Eventual plan to shift away from MATLAB to write code.

More random stuff

Turing completeness vs non turing complete languages for evolution.
Generative testing + Lexicase selection.

1 thought on “15 January 2014

  1. Pingback: Past Meeting Notes | Computational Intelligence Laboratory

Leave a Reply

Your email address will not be published. Required fields are marked *