Contrastive ECOC: Learning Output Codes for Adversarial Defense

August 14, 2025 · Entered Twilight · 🏛 International Conference on Information and Knowledge Management

Repo contents: AdversarialAttackCleverHans.py, AdversarialAttackCleverHans_ecoc.py, Codebooks, Dockerfile, README.md, acl_cfpc_finetune_testing.py, acl_cfpc_finetune_training.py, acl_finetune_testing.py, acl_finetune_training.py, acl_tfc_testing.py, acl_tfc_training.py, cleverhans, data_aug, exceptions, lars.py, models, run.py, simclr.py, simclr_finetune_testing.py, simclr_finetune_training.py, simple_testing.py, simple_training.py, utils.py

Authors Che-Yu Chou, Hung-Hsuan Chen arXiv ID 2508.10491 Category cs.LG: Machine Learning Cross-listed cs.AI, cs.IT Citations 0 Venue International Conference on Information and Knowledge Management Repository https://github.com/YuChou20/Automated-Codebook-Learning-with-Error-Correcting-Output-Code-Technique ⭐ 1 Last Checked 1 month ago

Abstract

Although one-hot encoding is commonly used for multiclass classification, it is not always the most effective encoding mechanism. Error Correcting Output Codes (ECOC) address multiclass classification by mapping each class to a unique codeword used as a label. Traditional ECOC methods rely on manually designed or randomly generated codebooks, which are labor-intensive and may yield suboptimal, dataset-agnostic results. This paper introduces three models for automated codebook learning based on contrastive learning, allowing codebooks to be learned directly and adaptively from data. Across four datasets, our proposed models demonstrate superior robustness to adversarial attacks compared to two baselines. The source is available at https://github.com/YuChou20/Automated-Codebook-Learning-with-Error-Correcting-Output-Code-Technique.