OOPS vs Lsearch
Lsearch: one problem; OOPS: problem sequence
OOPS: each time a new problem is solved, freeze (append) solution-computing program in readable, non-writeable storage. New programs may copy-edit / invoke previous code
Execute programs while generating them, such that prefixes may rewrite distribution on suffixes, using previous code.
New problem: spend half the search time on most recent frozen program and its prolongations, half the time on fresh programs starting after top frozen address. For universal solvers of all tasks: time-share second half (but not first!) on all tasks so far.
OOPS on real computers: complex, time-consuming instructions and efficient storage resets through optimal backtracking
8-bias-optimal way of reusing previous solutions! If there is a way of profiting from previous tasks then OOPS will find it.
Back to J. Schmidhuber's OOPS page