Usage

Use Pick n Mix to create a regression model:

from picknmix import Layer, Stack

import numpy as np
from sklearn.linear_model import LinearRegression
from sklearn.linear_model import Ridge

first_layer = Layer([LinearRegression(), Ridge()])
second_layer = Layer([LinearRegression()])
model = Stack([first_layer, second_layer])

X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])
y = np.dot(X, np.array([1, 2])) + 3
model.fit(X, y)
model.predict(np.array([[3, 5]]))

You can also use preprocessing in a Layer:

from sklearn.linear_model import MinMaxScaler

first_layer = Layer([LinearRegression(), Ridge()],
                    preprocessors = [MinMaxScaler(), None])

Classification task can also be done:

from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier

first_layer = Layer([LogisticRegression(solver='liblinear'),
                     RandomForestClassifier()],
                     proba=True)
second_layer = Layer([LogisticRegression(solver='liblinear')])
model = Stack([first_layer, second_layer])

X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])
y = np.array([1, 1, 0, 0])
model.fit(X, y)
model.predict(np.array([[1, 3]]))

You can also make use of cross-validation with one of the different scikit-learn options (e.g. Stratified K-Fold):

from sklearn import datasets
from sklearn.model_selection import StratifiedKFold

digits = datasets.load_digits()
features = digits.images
features = features.reshape(features.shape[0],
                            features.shape[1]*features.shape[2])
targets = digits.target
skf = StratifiedKFold(n_splits=2)
skf.get_n_splits(features, targets)

first_layer = Layer([LogisticRegression(solver='liblinear'),
                     RandomForestClassifier()],
                     proba=True)
second_layer = Layer([LogisticRegression(solver='liblinear')])
model = Stack([first_layer, second_layer], folds=skf)

model.fit(features, targets)