Awesome

Look-Ahead active learning with Neural Tangent Kernels

This repository contains an implementation of the proposed method for the paper "Making Look-Ahead Active Learning Strategies Feasible with Neural Tangent Kernels", to appear at NeurIPS 2022 by Mohamad Amin Mohamadi*, Wonhoe Bae* and Danica J. Sutherland.

Configs

Configs are written in the form of yaml. Please refer to the configs/cifar_resnet18.yml for an exmaple config and for the details about how to structure configs.

Datasets

Datasets are supposed to be located in data root which can be modified in the data section of a config file.

Train

To run the proposed method with the desired configuraiton, please refer to the command below.

python run.py --config_path configs/cifar_resnet18.yml --save_dir /tmp