Slide 1
Slide 1 text
Commodity Machine Learning
Past, present and future
Andreas Mueller
Slide 2
Slide 2 text
What is machine learning?
Slide 3
Slide 3 text
Automatic Decision Making
Spam?
Yes No
Slide 4
Slide 4 text
Spam?
Yes No
Slide 5
Slide 5 text
Programming
Machine Learning
Slide 6
Slide 6 text
Machine learning is EVERYWHERE
Slide 7
Slide 7 text
No content
Slide 8
Slide 8 text
No content
Slide 9
Slide 9 text
No content
Slide 10
Slide 10 text
Science
Engineering
Medicine
...
Slide 11
Slide 11 text
Commodity machine learning
Slide 12
Slide 12 text
past
Slide 13
Slide 13 text
+
Slide 14
Slide 14 text
No content
Slide 15
Slide 15 text
dawn of open source tools...
Slide 16
Slide 16 text
The age of shell
Slide 17
Slide 17 text
Documentation? Testing?
Slide 18
Slide 18 text
Scikit-learn: User centric machine learning
Slide 19
Slide 19 text
.fit(X, y)
.predict(X)
.transform(X)
Slide 20
Slide 20 text
present
Slide 21
Slide 21 text
Choose your ecosystem.
Slide 22
Slide 22 text
Open! Documented! Tested!
Slide 23
Slide 23 text
Usability is key!
Slide 24
Slide 24 text
ML Frameworks
PyMC, Edward, Stan
theano, tensorflow, keras
Slide 25
Slide 25 text
No content
Slide 26
Slide 26 text
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import Pipeline
Slide 27
Slide 27 text
github.com/scikitlearncontrib/scikitlearncontrib
Slide 28
Slide 28 text
(near) Future
Slide 29
Slide 29 text
pip install scikitlearn==0.18rc2
0.18
for the release candidate:
Slide 30
Slide 30 text
sklearn.cross_validation
sklearn.grid_search
sklearn.learning_curve
sklearn.model_selection
Slide 31
Slide 31 text
results = pd.DataFrame(grid_search.results_)
Slide 32
Slide 32 text
labels → groups
n_folds → n_splits
Slide 33
Slide 33 text
from sklearn.cross_validation import KFold
cv = KFold(n_samples, n_folds)
for train, test in cv:
...
from sklearn.model_selection import KFold
cv = KFold(n_folds)
for train, test in cv.split(X, y):
...
Slide 34
Slide 34 text
from sklearn.mixture import GaussianMixture
from sklearn.mixture import BayesianGaussianMixture
Slide 35
Slide 35 text
PCA()
RandomizedPCA()
PCA()
Slide 36
Slide 36 text
Gaussian Process Rewrite
Slide 37
Slide 37 text
Isolation Forests
Slide 38
Slide 38 text
Play
from sklearn.neural_network import MLPClassifier
Work
import keras
Slide 39
Slide 39 text
pipe = Pipeline([('preprocessing', StandardScaler()),
('classifier', SVC())])
param_grid = {'preprocessing': [StandardScaler(), None]}
grid = GridSearchCV(pipe, param_grid)
Slide 40
Slide 40 text
40
Slide 41
Slide 41 text
(further) Future
Slide 42
Slide 42 text
Feature / Column names
Slide 43
Slide 43 text
from __future__ import sklearn.plotting
Slide 44
Slide 44 text
from __future__ import AutoClassifier
Slide 45
Slide 45 text
More Transparency