/H-embedding_guided_hypernet-for-review-

Primary LanguagePython

H-embedding guided Hypernet

Setup Python Environment

  • Use conda to build environment:
$ conda env create -f environment.yml
$ conda activate hypercl_env
  • Set paths to your working directory and data directory correctly

Run experiment for Cifar10/100

  • Our method:

    Simply run the run_resnet.sh file under the cifar directory. The arguments are now set to conform with the results in paper, you may modify them as you wish.

  • Baselines:

    Currently we only support the ablation baselines, which could be derived respectively by changing the --emb_metric argument from Hembedding to random, and disabling --emb_reg