rosefun/weaklysupervisedlearning

Weakly supervised learning framework for classification.

Weakly Supervised Learning

Description

This is a Python framework for weakly supervised learning. This package can be used for weak supervised learning classification tasks.

Installation

You can install it by pip method.

pip install weaklysupervised

API

We have implemented following weakly-supervised learning algorithm.

• bootstrapping

Examples

from examples_utils import get_data, DNN
from sklearn import metrics
from weaklysupervised import BootstrappingNeuralNetworkClassifier

if __name__ == "__main__":
	X_train, X_test, y_train, y_test = get_data()
	DNN = DNN()
	clf = DNN.build_model(input_dim=30, output_dim=2)
	model = BootstrappingNeuralNetworkClassifier(clf, batch_size=128, epochs=40, bootstrapping_type="soft", 
	beta=0.95, patience=5, best_model_name="model_check_point_best_model")
	model.fit(X_train, y_train, validation_data=(X_test, y_test), )
	predict = model.predict(X_test)
	acc = metrics.accuracy_score(y_test, predict)
	print("bootstrapping accuracy", acc)

Please see examples folder for more examples.