Small tactical todos are now files as issues - this page now lists only more general todos of strategic importance

Short-term

• Extend the current trivial hillclimber to encompass lists, tuples, and functions
• Extend the plop minilanguage to action-perception and enum types. Think more carefully about side-effects.
• Implement a non-trivial learning algorithm
• Implement nontrivial representation-building for "flat" knobs
• Implemented deme management
• Run some interesting benchmarks on Boolean & real-valued & mixed datasets

Long-term

• Implement procedural abstraction via datamining & via direct transformation
• Primitive-recursive operations and associated transformations via either fold or explicit recursion
• Run benchmarks on the full set of benchmarks
• Exploiting uncertain background knowledge, learning coupling rules, metalearning