YoonyeongKim / SAAL

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

SAAL: Sharpness-Aware Active Learning (ICML 2023)

Requirements

To install requirements:

pip install -r requirements.txt

Training

To train the model(s) in the paper, run this command:

python3 main.py --data Cifar10 --acqMode Max_Diversity --optimizer_name Adam --isInit True
python3 main.py --data Cifar10 --acqMode Max_Diversity --optimizer_name SAM --isInit False

Evaluation

  • Data will be downloaded to folder 'data'.
  • Result will be recorded to folder 'Results'.

Results

Our model achieves the following performance on active learning settings:

Optimizer FashionMNIST SVHN CIFAR-10 CIFAR-100
Adam 85.8% 76.8% 54.4% 47.6%
SAM 86.3% 78.8% 57.0% 48.4%

About

License:Apache License 2.0


Languages

Language:Python 100.0%