MSD: Saliency-aware Knowledge Distillation for Multimodal Understanding

Woojeong Jin, Maziar Sanjabi, Shaoliang Nie, Liang Tan, Xiang Ren, Hamed Firooz


Abstract
To reduce a model size but retain performance, we often rely on knowledge distillation (KD) which transfers knowledge from a large “teacher” model to a smaller “student” model. However, KD on multimodal datasets such as vision-language tasks is relatively unexplored, and digesting multimodal information is challenging since different modalities present different types of information. In this paper, we perform a large-scale empirical study to investigate the importance and effects of each modality in knowledge distillation. Furthermore, we introduce a multimodal knowledge distillation framework, modality-specific distillation (MSD), to transfer knowledge from a teacher on multimodal tasks by learning the teacher’s behavior within each modality. The idea aims at mimicking a teacher’s modality-specific predictions by introducing auxiliary loss terms for each modality. Furthermore, because each modality has different saliency for predictions, we define saliency scores for each modality and investigate saliency-based weighting schemes for the auxiliary losses. We further study a weight learning approach to learn the optimal weights on these loss terms. In our empirical analysis, we examine the saliency of each modality in KD, demonstrate the effectiveness of the weighting scheme in MSD, and show that it achieves better performance than KD on four multimodal datasets.
Anthology ID:
2021.findings-emnlp.302
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2021
Month:
November
Year:
2021
Address:
Punta Cana, Dominican Republic
Venues:
EMNLP | Findings
SIG:
SIGDAT
Publisher:
Association for Computational Linguistics
Note:
Pages:
3557–3569
Language:
URL:
https://aclanthology.org/2021.findings-emnlp.302
DOI:
10.18653/v1/2021.findings-emnlp.302
Bibkey:
Cite (ACL):
Woojeong Jin, Maziar Sanjabi, Shaoliang Nie, Liang Tan, Xiang Ren, and Hamed Firooz. 2021. MSD: Saliency-aware Knowledge Distillation for Multimodal Understanding. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 3557–3569, Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
MSD: Saliency-aware Knowledge Distillation for Multimodal Understanding (Jin et al., Findings 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.findings-emnlp.302.pdf
Data
Hateful MemesSNLI-VE