auto_ml
auto_ml copied to clipboard
Improvements week of nov 6
- [x] prediction_intervals complete
- [x] allow the user to specify just one interval
- [x] only train up the intervals that users actually ask for
- [x] figure out return values (probably a dataframe by default)
- [ ] add documentation
- [ ] add examples
- [x] add more tests?
- [ ] look into how r does this, and consider using their return format
Raghav's ask for a more basic training
-
[ ] primarily, taking in the user's number of trees if they give it to us, rather than using X_test
-
[x] dataframevectorizer performance improvements
-
[x] we've already got an issue open for this- look there for more info.
- [x] remove tensorflow as a requirement. consider removing keras as a requirement as well? that might break things
- [ ] pin versions