Wycliffe's provided some advanced code (including the user of Pipline) for a classifier method here: What are the numbers in factor download? - #3 by Whycliffes
Thank you Wycliffe's. This is essential open-source code where Whycliffe's freely shares his ideas and welcomes any suggestions, I think. AND importantly, welcomes any additional code that Whycliffe's can look at and use at this discretion. Ultimately he can add any new ideas he likes to his code with modifications….which others can add to…which Whycliffe's can look at., others can modify,…which Whycliffe's look at…. It works elsewhere or that is my impression as a non-programmer. Pretty sure XGBoost uses this and has become pretty popular with a pretty good set of code. I like Sklearn. I am told Pandas is open-source.
Thank you again Whycliffe's! Well done.
I wonder if P123 would share the code it uses for Extra Trees Regresor, XGBoost, Ridge Regression and the rest.
It is mostly open-source with origins from Sklearn and XGBoost isn't it?
I guess you could have a beta-test with a few select users making comments that P123 may or may not incorporate ultimately. Even then wouldn't it be nice to have those most advanced at machine learning look at the code they are discussing?
Even if there are reasons to keep it a secret most of it is not really a secret ExtraTreesRegressor and K-fold validation, for example, are well documented and did not originate from anybody in the P123 forum. There is a limit to how much people will pay for open-source programming. P123 servers are a true value at the right cost. A cost that is not inelastic with people easily able to obtain alternative sources of coding and hardware. I do not pretend to be in charge of setting prices at P123 but Whycliffe's is essentially looking for an alternative source making the price elastic elastic for him, in an economic sense. Python code, being open source, is not a monopoly.
And Whyciffe's will ultimately succeed—using classifications as he wishes with his own code--based on what he has shared as a sample of his early (first?) machine learning programming. Again, thank you Whycliffe's.
Jim