An introduction to classification with pySkyNet¶

Classification : Prediction class membership for a new example.

In the following example we will use the SkyNetClassifier class to classify flowers in the iris dataset.

You will need sklearn library to load the iris dataset.

from sklearn.datasets import load_iris
from sklearn.utils import shuffle
from SkyNet import SkyNetClassifier

# X are the features and y are the targets
# shuffle returns a random permutation
X_class,y_class = shuffle(load_iris().data, load_iris().target)

We split the data into train,valid and test. The neural net will adjust the weights solely based on the train data. It will monitor the error on the validation data and stop the training once the error on the validation data fails to diminish. After that we calculate the predictions for the test data using the trained neural network.

X_train = X_class[0:70]
y_train = y_class[0:70]

X_valid = X_class[70:100]
y_valid = y_class[70:100]

X_test =X_class[100:]
y_test =y_class[100:]

We instantiate the neural network with 3 hidden layers with each 10 nodes (10, 10, 10) on the training data. With linear rectified units as activation functions for the hidden layers and linear activation for the outputlayer (3, 3, 3, 0) on a single core n_jobs=1. We use the string 'identification' as id.

sn_cla = SkyNetClassifier(id = 'identification_cla', n_jobs = 1, activation = (3, 3, 3, 0), layers = [10, 10, 10], max_iter = 200)

Now we perform the actual training of the neural network

sn_cla.fit(X_train, y_train, X_valid, y_valid)

Whereafter sn_cla.train_pred are the learned class probabilities for the training set.

Whereafter sn_cla.valid_pred are the learned class probabilities for for the validation set.

Getting the predictions for the test set

test_yhat = sn_cla.predict_proba(X_test)

All code combined

from sklearn.datasets import load_iris
from sklearn.utils import shuffle
from SkyNet import SkyNetClassifier

# X are the features and y are the targets
# shuffle returns a random permutation
X_class,y_class = shuffle(load_iris().data, load_iris().target)

X_train = X_class[0:70]
y_train = y_class[0:70]

X_valid = X_class[70:100]
y_valid = y_class[70:100]

X_test =X_class[100:]
y_test =y_class[100:]

sn_cla = SkyNetClassifier(id = 'identification_cla', n_jobs = 1, activation = (3, 3, 3, 0), layers = (10, 10, 10), max_iter = 200)

sn_cla.fit(X_train, y_train, X_valid, y_valid)

test_yhat = sn_cla.predict_proba(X_test)