Pedro Ribeiro
Pedro Ribeiro
In base evolver, it is common to have idle cores while the last few individuals in a population are being evaluated. In the steady state evolver, a new individual is...
Steps 1. write a function that converts an sklearn pipeline into a graphpipeline. 2. Another function that converts a graphpipeline into a graphpipelineIndividual. 3. An individual generator that takes in...
Add a function that takes in a string name for a sklearn pipeline or the sklearn type (e.g "LogisticRegression" OR sklearn.linear_model.LogisticRegression) and returns the appropriate params function. So that we...
TPOT1 had some files with licensing text from other packages (for example: https://github.com/EpistasisLab/tpot/blob/master/tpot/builtins/one_hot_encoder.py) Do we add those here as well? Additionally, TPOT1 started its source files with the following text....
Currently, these are built into the evolver class. If we merge the revolver classes, this may be easier to use as a standalone function #65 The two strategies for pruning...
TPOT can get very good at overfitting the cv score. Maybe shuffling the cv fold every generation would counteract this? It would add noise to the learning process, but in...
TPOT2 can often get stuck without learning for long periods before finding a "breakthrough." One idea to potentially try to break out of this sooner could be to increase the...
It would loop through the leaves (selectors), compile a list of all unique selected features, and then return the length of that list.