addeed policy learning to readme #440
Conversation
README.md
Outdated
| ``` | ||
|
|
||
|
|
||
|  |
There was a problem hiding this comment.
Why put the image under notebooks?
There was a problem hiding this comment.
Not that it's super important, but changes to the notebooks directory trigger the notebook tests to run, which should not be necessary for this change.
(However, if you create a new directory (/images/, say) that we want to allow changes to without running tests you'll need to add a line for it here; the change to the pipelines file will itself force all tests to run, but then subsequent changes to that directory will be free)
There was a problem hiding this comment.
that's were all our images are
There was a problem hiding this comment.
I'll move then all readme images there. That should still trigger notebook tests, as I'll need to move images out of the notebooks folder
There was a problem hiding this comment.
Created a new image folder and moved only that image there. Added that image fodler to azure-pielines
| from sklearn.ensemble import RandomForestRegressor | ||
|
|
||
| # fit a single binary decision tree policy | ||
| policy = DRPolicyTree(max_depth=1, min_impurity_decrease=0.01, honest=True) |
There was a problem hiding this comment.
Are the variations from the default init args important for this example? Would a user understand why to set these particular values?
Or in other words, if I'm going to copy and paste this to start playing around with it, are these reasonable settings, and if so why are they not already the defaults?
There was a problem hiding this comment.
not sure. migth be reasonable. Was more showing some capabilities of what is parameterizable
There was a problem hiding this comment.
If the defaults are reasonable, and these arguments are also reasonable, and there's no particular reason to assume one is better than the other, then that's fine.
Otherwise, I'd try to make sure that whatever you put here will be a good start for someone to copy and paste, rather than show off functionality that's more advanced but requires more knowledge to use well.
There was a problem hiding this comment.
yeap I think this is good and along these lines
| policy.feature_importances_ | ||
|
|
||
| # fit a binary decision forest | ||
| policy = DRPolicyForest(max_depth=1, min_impurity_decrease=0.01, honest=True) |
There was a problem hiding this comment.
Same question about defaults here.
There was a problem hiding this comment.
yeah same. seems good parameters the users might want to play with.
No description provided.