Realistic OOPS applications
Problem sequence: E.g., n-th problem is: find a shorter path than best so far. Or predict events 1,..,n.
Bias: task-dependent distribution on programs built from primitives (e.g., probabilistic syntax diagram)
Primitives can be anything: Theorem provers, neural net algorithms, trajectory generators for robot arms, SVMs, Bayesian network algorithms, assembler-like instructions, your favorite search and learning algorithms, combinations thereof...
Just make sure OOPS can interrupt any primitive once it consumes too much time as part of a program composed and tested by OOPS!
Back to J. Schmidhuber's OOPS page