Summary of Discom-kd: Cross-modal Knowledge Distillation Via Disentanglement Representation and Adversarial Learning, by Dino Ienco (evergreen et al.
DisCoM-KD: Cross-Modal Knowledge Distillation via Disentanglement Representation and Adversarial Learning
by Dino Ienco, Cassio Fraga Dantas
First submitted to arxiv on: 5 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel approach to cross-modal knowledge distillation, dubbed DisCoM-KD, is proposed, departing from the traditional teacher/student paradigm. This framework combines disentanglement representation learning with adversarial domain adaptation to extract modality-specific features for a specific downstream task. Unlike traditional methods, DisCoM-KD simultaneously learns all single-modal classifiers, eliminating the need to learn each student model separately as well as the teacher classifier. The effectiveness of DisCoM-KD is demonstrated on three standard multi-modal benchmarks, outperforming recent state-of-the-art knowledge distillation frameworks in mismatch scenarios involving both overlapping and non-overlapping modalities. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary DisCoM-KD is a new way to share knowledge between different types of data. Instead of using a teacher to teach students, this method learns multiple student models at the same time. It does this by combining two techniques: disentanglement representation learning and adversarial domain adaptation. This allows DisCoM-KD to extract important features from each type of data. The results show that DisCoM-KD is better than other methods at sharing knowledge between different types of data. |
Keywords
» Artificial intelligence » Domain adaptation » Knowledge distillation » Multi modal » Representation learning » Student model
