ypeleg / hungabunga Goto Github PK

View Code? Open in Web Editor NEW

707.0 23.0 79.0 297 KB

HungaBunga: Brute-Force all sklearn models with all parameters using .fit .predict!

License: MIT License

Python 100.00%

scikit-learn brute force machine learning kaggle automl sklearn fit predict

hungabunga's Introduction

Hunga-Bunga

Brute Force all scikit-learn models and all scikit-learn parameters with fit predict.

Lets brute force all sklearn models with all of sklearn parameters! Ahhh Hunga Bunga!!

from hunga_bunga import HungaBungaClassifier, HungaBungaRegressor

And then simply:

What?

Yes.

No! Really! What?

Many believe that

most of the work of supervised (non-deep) Machine Learning lies in feature engineering, whereas the model-selection process is just running through all the models or just take xgboost.

So here is an automation for that.

HOW IT WORKS

Runs through all sklearn models (both classification and regression), with all possible hyperparameters, and rank using cross-validation.

MODELS

Runs all the model available on sklearn for supervised learning here. The categories are:

Generalized Linear Models
Kernel Ridge
Support Vector Machines
Nearest Neighbors
Gaussian Processes
Naive Bayes
Trees
Neural Networks
Ensemble methods

Note: Some models were dropped out (nearly none of them..) and some crash or cause exceptions from time to time. It takes REALLY long to test this out so clearing exceptions took me a while.

Installation

pip install hunga-bunga

Dependencies


- Python (>= 2.7)
- NumPy (>= 1.11.0)
- SciPy (>= 0.17.0)
- joblib (>= 0.11)
- scikit-learn (>=0.20.0)
- tabulate (>=0.8.2)
- tqdm (>=4.28.1)

Option I (Recommended): brain = False

As any other sklearn model

clf = HungaBungaClassifier()
clf.fit(x, y)
clf.predict(x)

And import from here

from hunga_bunga import HungaBungaClassifier, HungaBungaRegressor

Option II: brain = True

As any other sklearn model

clf = HungaBungaClassifier(brain=True)
clf.fit(x, y)

The output looks this:

Model	accuracy	Time/clf (s)
SGDClassifier	0.967	0.001
LogisticRegression	0.940	0.001
Perceptron	0.900	0.001
PassiveAggressiveClassifier	0.967	0.001
MLPClassifier	0.827	0.018
KMeans	0.580	0.010
KNeighborsClassifier	0.960	0.000
NearestCentroid	0.933	0.000
RadiusNeighborsClassifier	0.927	0.000
SVC	0.960	0.000
NuSVC	0.980	0.001
LinearSVC	0.940	0.005
RandomForestClassifier	0.980	0.015
DecisionTreeClassifier	0.960	0.000
ExtraTreesClassifier	0.993	0.002

The winner is: ExtraTreesClassifier with score 0.993.

hungabunga's People

Contributors

Stargazers

Watchers

Forkers

spectra456 deetanshu gzoumpourlis sanjitjain2 bhaweshchandola mmann1123 lingyu001 adewin ml-tools foeinlove trendingtechnology teosoft7 ghostintheshellarise arita37 hhy5277 shaunstanislauslau lapa34 satishcse sprinterzzj roidangur nirupam1sharma alliswellbro thomascherickal santhu45482 shalevy1 forkdump srikanthvadapalli madhavadama chandansinha oeeckhoutte tekmen0 suifengyougu veshitala chrisby iggisv9t rishi-wqd190004 meetrahulverma navaneeshgangala fervillarce indra-ji vishalbelsare dushyanttara denizkavi abhinavjain13 aakashofficial amirstudy heshenxian1 adillibabu vamsikavuru shlomihod awoziji laranea morkarslaanesh tilfast yumbohorquez renatacgcastanha liranbz romanzavodskikh amir9979 billy-horn 2d1ff1cult arpitartworks tsabarm drawdeveloper ryanxju isaaccheong95 akupitz adahmed-coursera mr-mad-phyco bruno81930 iq-scm barvin04 abeusher parasyteunknown thegoldenchicken shatz01 hamzaeialaoui

hungabunga's Issues

parameter ranges?

Hey! How did you determine the parameter ranges for all the models?
And have you thought about the implications of having them specified by name and not by model (i.e. alpha might be very different things for different models).

Documentation issue

HOW IT WORKS

BUT DOES IT WORK THO

JK this was funny and cool though. Thanks for the laugh :)

maybe also add scaling and normalizing...

hehe, nice repo...

i faced a similar challenge (and approach ;)) when fusing results of my graph pattern learner, but as it's part of a bigger thing, i rewrote many things to use scoop (instead of joblib).

Might still be worth a look for you, as it also adds standard scaling and normalizing as pipelines into the mix: https://github.com/RDFLib/graph-pattern-learner/blob/master/fusion/trained.py#L204

Hanging up on NuSVC

I'm not sure if this is a real issue or just the computer i'm running on being too slow, but, every time I try to run example.py, it hangs up with the loading bar at 67%.
I added brain=True for a more verbose output and saw that it hangs up after displaying "NuSVC".
For comparison, it takes me 30s-1m to get to that point, and I tried letting it run overnight and it still didn't get further than that... I killed it after maybe 24 hours because I was tired of having 7 cores out of 8 permanently maxxed.

Module not found error

getting this error:
"
4 from multiprocessing import cpu_count
5 from sklearn.base import BaseEstimator
----> 6 from regression import HungaBungaRegressor
7 from classification import HungaBungaClassifier
8
ImportError: No module named 'regression'
"
Can anyone please help in this issue

Please add gradient boosting

Please, add gradient boosting libraries, like CatBoost, XGBoost, LightGBM.

Multi-label classification

If I want multi-label, for the y vector should I use OHC or a 1-d array?

No module named 'regression'

Hi,
I am getting below error while initializing. Please advise the solution

FYI - All libraries are updated with latest version

ModuleNotFoundError Traceback (most recent call last)
in ()
----> 1 from hunga_bunga import HungaBungaClassifier, HungaBungaRegressor, HungaBungaZeroKnowledge

~\Anaconda3\lib\site-packages\hunga_bunga_init_.py in ()
4 from multiprocessing import cpu_count
5 from sklearn.base import BaseEstimator
----> 6 from regression import HungaBungaRegressor
7 from classification import HungaBungaClassifier
8

ModuleNotFoundError: No module named 'regression'

    X_train, y_train = prepare_data_for_ml(X_train, y_train)
    y_train, y_test = prepare_data_for_ml(X_test, y_test)

    clf = HungaBungaClassifier()
    clf.fit(X_train, y_train)

    clf.predict(X_test)

That's the error:

100%|██████████| 15/15 [00:00<00:00, 43.02it/s]
Traceback (most recent call last):
  File "/Users/yonatab/PycharmProjects/VisualFitnessUtils/activate_ml/activate_ml_on_joints.py", line 62, in <module>
    main()
  File "/Users/yonatab/PycharmProjects/VisualFitnessUtils/activate_ml/activate_ml_on_joints.py", line 43, in main
    clf.predict(X_test)
  File "/Users/yonatab/opt/anaconda3/envs/CondaEnv/lib/python3.7/site-packages/hunga_bunga/classification.py", line 202, in predict
    return self.model.predict(x)
  File "/Users/yonatab/opt/anaconda3/envs/CondaEnv/lib/python3.7/site-packages/sklearn/linear_model/base.py", line 289, in predict
    scores = self.decision_function(X)
  File "/Users/yonatab/opt/anaconda3/envs/CondaEnv/lib/python3.7/site-packages/sklearn/linear_model/base.py", line 263, in decision_function
    "yet" % {'name': type(self).__name__})
sklearn.exceptions.NotFittedError: This SGDClassifier instance is not fitted yet

Am I missing something?

Thanks

get all inf?

ExtraTreesClassifier

Model accuracy Time/clf (s)

SGDClassifier -inf inf
LogisticRegression -inf inf
Perceptron -inf inf
PassiveAggressiveClassifier -inf inf
MLPClassifier -inf inf
KMeans -inf inf
KNeighborsClassifier -inf inf
NearestCentroid -inf inf
RadiusNeighborsClassifier -inf inf
SVC -inf inf
NuSVC -inf inf
LinearSVC -inf inf
RandomForestClassifier -inf inf
DecisionTreeClassifier -inf inf
ExtraTreesClassifier -inf inf

The winner is: SGDClassifier with score -inf.

I used the same data with just RandomForestClassifier and got a [bad but] working classifier...

Error in initial import

Hi,
I downloaded and put the package in site-packages.
When importing, I get the error:


    from regression import HungaBungaRegressor
ModuleNotFoundError: No module named 'regression'

ypeleg / hungabunga Goto Github PK

hungabunga's Introduction

Hunga-Bunga

Lets brute force all sklearn models with all of sklearn parameters! Ahhh Hunga Bunga!!

And then simply:

What?

No! Really! What?

HOW IT WORKS

MODELS

Installation

Option I (Recommended): brain = False

Option II: brain = True

hungabunga's People

Contributors

Stargazers

Watchers

Forkers

hungabunga's Issues

ExtraTreesClassifier

Recommend Projects

Recommend Topics

Recommend Org