zeef 0.1.2

A Python Framework for Deep Active Learning

Zeef: Active Learning for Data-Centric AI

An active learning framework that can be applied to real-world scenarios that leak labeled data.

Installation

pip install zeef

For the local development, you can install from the Anaconda environment by

conda env create -f environment.yml

Quick Start

We can start from the easiest example: random select data points from an unlabeled data pool.

from sklearn import svm

from zeef.data import Pool
from zeef.learner.sklearn import Learner
from zeef.strategy import RandomSampling

learner = Learner(net=svm.SVC(probability=True))  # define the learner.
data_pool = Pool(unlabeled_data)  # generate the data pool.
strategy = RandomSampling(data_pool, learner=learner)  # define the sampling strategy.

query_ids = strategy.query(1000)  # query 1k samples for labeling.
strategy.update(query_ids, data_labels)  # label the 1k samples.
strategy.learn()  # train the model using all the labeled data.
strategy.infer(test_data)  # evaluate the model.

A quick MNIST CNN example can be found in here. Run

python main_torch.py

to start the quick demonstration.

License

Apache License 2.0