ClimbsRocks auto_ml issues

ClimbsRocks / auto_ml

[UNMAINTAINED] Automated machine learning for analytics & production

http://auto-ml.readthedocs.io

MIT License

1.64k stars 310 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

predict intervals master list

#292 ClimbsRocks opened 7 years ago
0
write tests for predict_intervals!

#291 ClimbsRocks opened 7 years ago
0
API for predict_intervals

#290 ClimbsRocks opened 7 years ago
0
add some regularization to gradientboosting warm starting

#289 ClimbsRocks opened 7 years ago
0
use quantile with a 0.5 alpha to get a prediction of the median

#288 ClimbsRocks opened 7 years ago
1
do model training differently inside gscv

#287 ClimbsRocks opened 7 years ago
0
only pass in training_params for final model, not for feature_learning

#286 ClimbsRocks opened 7 years ago
0
let the user pass in their own eval data set

#285 ClimbsRocks opened 7 years ago
0
adjust verbosity of sequential training during grid searching

#284 ClimbsRocks opened 7 years ago
0
naming considerations

#283 ClimbsRocks opened 7 years ago
0
have something like "is_hyperparameter_search" for eascv finalmodelatc issues

#282 ClimbsRocks opened 7 years ago
0
oversampling- move feature_learning to after oversampling

#281 ClimbsRocks opened 7 years ago
0
oversampling- overfitting

#280 ClimbsRocks opened 7 years ago
0
oversampling- consider a true holdout set

#279 ClimbsRocks opened 7 years ago
0
create an oversampling module

#278 ClimbsRocks opened 7 years ago
0
only use the original training data as validation data when using oversampling

#277 ClimbsRocks opened 7 years ago
0
give the user the option of passing in a function that's called after transform, and right before predict?

#276 ClimbsRocks opened 7 years ago
0
oversampling- use only original data

#275 ClimbsRocks opened 7 years ago
0
add an option to take in an already transformed matrix

#274 ClimbsRocks opened 7 years ago
0
add a .transform_only option

#273 ClimbsRocks opened 7 years ago
0
create advanced scoring logging for multi-class classification

#272 ClimbsRocks opened 7 years ago
1
lightgbm seems prone to overfitting with warm_start

#271 ClimbsRocks opened 7 years ago
0
eventually, train on the entire dataset even when warm_starting, or early_stopping

#270 ClimbsRocks opened 7 years ago
0
allow keyboard interrupt for any warm-started model

#269 ClimbsRocks opened 7 years ago
0
warm start for xgb

#268 ClimbsRocks opened 7 years ago
0
warm start for lightgbm

#267 ClimbsRocks opened 7 years ago
0
automatically pick theano as the backend for windows, and remove tf from requirements.txt??

#266 ClimbsRocks opened 7 years ago
0
tf takes forever to import- let the user specify which backend they want through an environmental variable

#265 ClimbsRocks opened 7 years ago
0
New ML Idea: use many basic model types for gradient boosting, not just decision trees

#264 ClimbsRocks opened 7 years ago
0
New ML Idea: learning rate schedules for gradient boosting trees

#263 ClimbsRocks opened 7 years ago
0
Build gradient boosting trees sequentially, with a smaller learning rate

#262 ClimbsRocks closed 7 years ago
1
SGD seems to be throwing an error?

#261 ClimbsRocks opened 7 years ago
0
Load checkpointed model

#260 ClimbsRocks opened 7 years ago
0
Evolving

#259 ClimbsRocks closed 7 years ago
1
sort printed params by score

#258 ClimbsRocks closed 7 years ago
1
print out params as their own columns in the df

#257 ClimbsRocks closed 7 years ago
1
do post-hoc feature selection, to tell the user which features they can cut

#256 ClimbsRocks opened 7 years ago
0
force deep learning to always do feature scaling

#255 ClimbsRocks closed 7 years ago
1
Have a `timeit` param?

#254 ClimbsRocks opened 7 years ago
0
deep learning optimizations to make

#253 ClimbsRocks opened 7 years ago
3
search printing: remove `model__` prefix from logging

#252 ClimbsRocks closed 7 years ago
1
Keras callbacks to add

#251 ClimbsRocks opened 7 years ago
2
Ensembling 3.0

#250 ClimbsRocks opened 7 years ago
0
don't report feature_responses for feature_learning

#249 ClimbsRocks closed 7 years ago
2
separate param for optimize_feature_learning_model

#248 ClimbsRocks closed 7 years ago
1
create feature_responses only for the top 100 features

#247 ClimbsRocks closed 7 years ago
2
use RMSE instead of MSE for keras reporting

#246 ClimbsRocks opened 7 years ago
0
Analytics improvements

#245 ClimbsRocks opened 7 years ago
0
for deep learning, sort feature_responses by magnitude of effect

#244 ClimbsRocks opened 7 years ago
1
add learning rate schedules for DL

#243 ClimbsRocks closed 7 years ago
1

Previous Next