Overview

ProgLearn provides classes and functions for biological machine learning. Notably, it improves in performance on all tasks (including past and future) with any new data. This sets it apart from classical machine learning algorithms and many other recent approaches to biological learning.

The Library

All classes and functions are available through the ProgLearn package and can also be imported separately.

import proglearn as PL
from proglearn.forest import UncertaintyForest

The Learners

There are three main parts to the ProgLearn package: Lifelong Classification Network, Lifelong Classification Forest, and the Uncertainty Forest. All three have very similar syntax and usage. A general overview is provided below with more specific and complete examples in the tutorial section. This overview example will use proglearn.forest.UncertaintyForest but is generalizable to the Lifelong Classification Network and Lifelong Classification Forest.

First, we'll create our forest:

UF = UncertaintyForest(n_estimators = n_estimators)

Then, we fit to data:

UF.fit(X_train, y_train)

Finally we can predict the classes of the data:

predictions = UF.predict(X_test)

Another advantage of the ProgLearn package is the predict_proba(X) function. It can be used to estimate the class posteriors for each example in the input data, X.

predicted_posteriors = UF.predict_proba(X_test)

Wrap Up

This overview covers the basics of using ProgLearn. Most use cases will utilize the functions presented above. Further examples are available in the tutorials section (see menu).