This repository has been archived by the owner. It is now read-only.
Permalink
Switch branches/tags
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
256 lines (179 sloc) 7.58 KB

dask-searchcv

Note

Dask-SearchCV is now included in Dask-ML <http://dask-ml.readthedocs.io/>_.

Further development to Dask-SearchCV is occuring in the Dask-ML <https://github.com/dask/dask-ml>_ repository. Please post issues and make pull requests there.

Tools for performing hyperparameter optimization of Scikit-Learn models using Dask.

Introduction

This library provides implementations of Scikit-Learn's GridSearchCV and RandomizedSearchCV. They implement many (but not all) of the same parameters, and should be a drop-in replacement for the subset that they do implement. For certain problems, these implementations can be more efficient than those in Scikit-Learn, as they can avoid expensive repeated computations.

For more information, see this blogpost.

Highlights

Install

Dask-searchcv is available via conda or pip:

# Install with conda
$ conda install dask-searchcv -c conda-forge

# Install with pip
$ pip install dask-searchcv

Walkthrough

Drop-In Replacement

Dask-searchcv provides (almost) drop-in replacements for Scikit-Learn's GridSearchCV and RandomizedSearchCV. With the exception of a few keyword arguments, the api's are exactly the same, and often only an import change is necessary:

Flexible Backends

Dask-searchcv can use any of the dask schedulers. By default the threaded scheduler is used, but this can easily be swapped out for the multiprocessing or distributed scheduler:

# Distribute grid-search across a cluster
from dask.distributed import Client
scheduler_address = '127.0.0.1:8786'
client = Client(scheduler_address)

search.fit(digits.data, digits.target)

Works Well With Dask Collections

Dask collections such as dask.array, dask.dataframe and dask.delayed can be passed to fit. This means you can use dask to do your data loading and preprocessing as well, allowing for a clean workflow. This also allows you to work with remote data on a cluster without ever having to pull it locally to your computer:

import dask.dataframe as dd

# Load data from s3
df = dd.read_csv('s3://bucket-name/my-data-*.csv')

# Do some preprocessing steps
df['x2'] = df.x - df.x.mean()
# ...

# Pass to fit without ever leaving the cluster
search.fit(df[['x', 'x2']], df['y'])

Avoid Repeated Work

When searching over composite estimators like sklearn.pipeline.Pipeline or sklearn.pipeline.FeatureUnion, dask-searchcv will avoid fitting the same estimator + parameter + data combination more than once. For pipelines with expensive early steps this can be faster, as repeated work is avoided.

For example, given the following 3-stage pipeline and grid (modified from this scikit-learn example).

from sklearn.feature_extraction.text import CountVectorizer, TfidfTransformer
from sklearn.linear_model import SGDClassifier
from sklearn.pipeline import Pipeline

pipeline = Pipeline([('vect', CountVectorizer()),
                     ('tfidf', TfidfTransformer()),
                     ('clf', SGDClassifier())])

grid = {'vect__ngram_range': [(1, 1)],
        'tfidf__norm': ['l1', 'l2'],
        'clf__alpha': [1e-3, 1e-4, 1e-5]}

the Scikit-Learn grid-search implementation looks something like (simplified):

scores = []
for ngram_range in parameters['vect__ngram_range']:
        for norm in parameters['tfidf__norm']:
                for alpha in parameters['clf__alpha']:
                        vect = CountVectorizer(ngram_range=ngram_range)
                        X2 = vect.fit_transform(X, y)
                        tfidf = TfidfTransformer(norm=norm)
                        X3 = tfidf.fit_transform(X2, y)
                        clf = SGDClassifier(alpha=alpha)
                        clf.fit(X3, y)
                        scores.append(clf.score(X3, y))
best = choose_best_parameters(scores, parameters)

As a directed acyclic graph, this might look like:

"scikit-learn grid-search directed acyclic graph"

In contrast, the dask version looks more like:

scores = []
for ngram_range in parameters['vect__ngram_range']:
        vect = CountVectorizer(ngram_range=ngram_range)
        X2 = vect.fit_transform(X, y)
        for norm in parameters['tfidf__norm']:
                tfidf = TfidfTransformer(norm=norm)
                X3 = tfidf.fit_transform(X2, y)
                for alpha in parameters['clf__alpha']:
                        clf = SGDClassifier(alpha=alpha)
                        clf.fit(X3, y)
                        scores.append(clf.score(X3, y))
best = choose_best_parameters(scores, parameters)

With a corresponding directed acyclic graph:

"dask-searchcv grid-search directed acyclic graph"

Looking closely, you can see that the Scikit-Learn version ends up fitting earlier steps in the pipeline multiple times with the same parameters and data. Due to the increased flexibility of Dask over Joblib, we're able to merge these tasks in the graph and only perform the fit step once for any parameter/data/estimator combination. For pipelines that have relatively expensive early steps, this can be a big win when performing a grid search.

Index

.. toctree::

    api