EIPsilly/BiasedAD

BiasedAD

environment

Python version 3.7.16.

Create suitable conda environment:

conda env create -f environment.yml

fashionMNIST

For the fashionMNIST dataset, our training data contains three categories: normal, non_target, and target, which need to be explicitly specified in the sh command.

$\text{BiasedAD}$

exec -a "BAD_420" python main.py --model_type BiasedAD --dir_path ./result/fmnist --dataset_name fashionmnist --normal_class 4 --non_target_outlier_class 2 --target_outlier_class 0 --gpu 2 --random_seed 0&

$\text{BiasedAD}^\text{M}$

$\text{BiasedAD}^\text{M}$ is similar to $\text{BiasedAD}$. Although the command includes --non_target_outlier_class, the number of non-target anomalies is set to 0 during runtime. The required sampling count is set to default 100 and does not need to be explicitly declared.

exec -a "BADM_420" python main.py --model_type BiasedADM --dir_path ./result/fmnist --dataset_name fashionmnist --normal_class 4 --non_target_outlier_class 2 --target_outlier_class 0 --gpu 2 --random_seed 0 &

nb-15

$\text{BiasedAD}$

exec -a "BAD_nb15" python main.py --model_type BiasedAD --dir_path ./result/nb15 --dataset_name nb15 --gpu 2 --random_seed 0&

$\text{BiasedAD}^\text{M}$

$\text{BiasedAD}^\text{M}$ adds the parameter --sample_count compared to $\text{BiasedAD}$. Since the default sampling count is 100, which is specific to the fashionMNIST dataset, it's necessary to declare --sample_count 1000

exec -a "BADM_nb15" python main.py --model_type BiasedADM --dir_path ./result/nb15 --dataset_name nb15 --gpu 0 --sample_count 1000 --random_seed 0 &