Skip to content
forked from JordanAsh/badge

An implementation of the BADGE batch active learning algorithm.

Notifications You must be signed in to change notification settings

AminParvaneh/badge

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Batch Active learning by Diverse Gradient Embeddings (BADGE)

An implementation of the BADGE batch active learning algorithm. Details are provided in our paper, Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds, which was presented as a talk in ICLR 2020. This code was built by modifying Kuan-Hao Huang's deep active learning repository.

Dependencies

To run this code fully, you'll need PyTorch (we're using version 1.4.0), scikit-learn, and OpenML. We've been running our code in Python 3.7.

Running an experiment

python run.py --model resnet --nQuery 1000 --data CIFAR10 --alg badge
runs an active learning experiment using a ResNet and CIFAR-10 data, querying batches of 1,000 samples according to the BADGE algorithm. This code allows you to also run each of the baseline algorithms used in our paper.

python run.py --model mlp --nQuery 10000 --did 6 --alg conf
runs an active learning experiment using an MLP and dataset number 6 from OpenML, querying batches of 10,000 with confidence sampling. Note that in our code, OpenML datasets can only be used with MLP architectures.

Analyzing experimental results

See the readme file in scripts/ for more details about generating plots like those in our paper.

About

An implementation of the BADGE batch active learning algorithm.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%