How to check the searched formula? about deep-symbolic-optimization HOT 3 CLOSED

Hey, thanks for your interest! The code currently saves only the final expression at the end of each training run. If you do a simple run with python -m dsr.run config.json it will save each run summary as a single line in benchmark_dsr.csv. The column "traversal" will show you the best found expression. The logs like dsr_Nguyen-1_0.csv show batch statistics each training step.

In a recent update, we've added two additional logging features: 1) saving the "hall of fame" (top N expressions found during all of training) and 2) saving the Pareto front (based on a complexity measure). Would this be what you're looking for?

Also please note we will push a new release version shortly, which will include these features.

from deep-symbolic-optimization.

Also, if you're interested in using the instantiated expression (e.g. as an executable function), this came up in #5 and support for this will also be added shortly.

from deep-symbolic-optimization.

Hi @pone7 , the requested feature has been added. You can now specify a "hall of fame" and/or Pareto front. If using the sklearn interface, you can also interact with the best expression via DeepSymbolicRegressor.program_.

from deep-symbolic-optimization.