Usage¶

PyLearn provides different features from supervised and unsupervised learning.

To use them, just import pylearn:

import pylearn as pl

Each model can be saved to storage to prevent training your model again. Just import it:

import pl.save, pl.load

If you want to normalize your input data, simply import:

import pl.min_max_normalization, pl.z_normalization

You can evaluate every model with accuracy, precision, recall and F1 score:

import pl.accuracy, pl.precision, pl.recall, pl.f1_score

Change numbers into a one hot representation:

import pl.to_one_hot

The major features are:

Classification¶

You can use Gaussian Naive Bayes.

Gaussian works for continuous data. Multinomial Naive Bayes works perfect for text classification and will come in version 1.1.0.

The usage is quite simple:

gnb = pl.GaussianNaiveBayes()

Now, train the model by using the fit function:

gnb.fit(features, output)

Let the model predict your input:

gnb.predict(features)

You can choose between K-Means and K-Medoids as clustering models.

The usage of both is quite similar:

kmeans = pl.KMeans()
kmedoids = pl.KMedoids()

Now, train the model by using the fit function, we will use kmeans to continue:

kmeans.fit(points)

This returns a list of the to the data points assigned clusters. You could visualize the result with matplotlib.

If you want to customize the result, the following functions may help you:

kmeans.assigned_clusters(any_cluster)
kmeans.rename(old, new)

The neural network comes with different activation functions and loss functions.

First, you need to create a network, for example:

network = [
     pl.Dense_layer(input_length, output_length),
     pl.Tanh(),
     plpDense_layer(input_length, output_length),
     pl.Tanh()
 ]

Now, train the model:

pl.NeuralNetwork.fit(x_train, y_train, network, loss, loss_derivative, epochs, log_error, log_duration)

Let the model predict your input:

pl.NeuralNetwork.predict(x, network)