• Currenty GPMaLMO optimises embedding accuracy and minimises tree size to make readable trees
• However tree size is a poor proxy for tree interpretability
• This is because it doesn't take into account functional complexity (among other things)
• We want to try to improve the complexity measurment so we can better optimise for tree readability

Done:

In Testing:

• Runtime optimisation
• Functional (dictionary) complexity

Abandoned:

• Tikhonov regularisation (smoothing)

Usage (from the src/ directory):
To run on iris.data (in /src/datasets dir) for 10 generations, using size minimisation as our seconday objective to minimise (alongside neighbourhood structure embedding loss)

make run DATASET=iris GENS=10 OBJ=size

To clean up output files in the directory:

make clean

To run (non exhaustive) tests

make test

Note:

• Datasets used in the paper are in datasets/
• Add your own datasets in csv format, with a header line:
  Header: classPosition,#features,#classes,seperator. e.g.
  classLast,1024,20,comma (from COIL20.data)
• Most GP parameters are configured in gpmalmo/rundata.py