Papers › Noise Estimation Using Density Estimation for Self-Supervised Multimodal Learning

Noise Estimation Using Density Estimation for Self-Supervised Multimodal Learning

6 Mar 2020arXiv:2003.03186archive 2025-07-28

Elad Amrani, Rami Ben-Ari, Daniel Rotman, Alex Bronstein

One of the key factors of enabling machine learning models to comprehend and solve real-world tasks is to leverage multimodal data. Unfortunately, annotation of multimodal data is challenging and expensive. Recently, self-supervised multimodal methods that combine vision and language were proposed to learn multimodal representations without annotation. However, these methods often choose to ignore the presence of high levels of noise and thus yield sub-optimal results. In this work, we show that the problem of noise estimation for multimodal data can be reduced to a multimodal density estimation task. Using multimodal density estimation, we propose a noise estimation building block for multimodal representation learning that is based strictly on the inherent correlation between different modalities. We demonstrate how our noise estimation can be broadly integrated and achieves comparable results to state-of-the-art performance on five different benchmark datasets for two challenging multimodal tasks: Video Question Answering and Text-To-Video Retrieval. Furthermore, we provide a theoretical probabilistic error bound substantiating our empirical results and analyze failure cases. Code: https://github.com/elad-amrani/ssml.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

elad-amrani/ssml officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Density EstimationNoise EstimationQuestion AnsweringRepresentation LearningRetrievalText to Video RetrievalVideo Question AnsweringVideo RetrievalVisual Question AnsweringVisual Question Answering (VQA)Zero-Shot Video Retrieval

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Video Retrieval MSVD SSML text-to-video Mean Rank -- #23 of 24 Archive leaderboard report
Video Retrieval MSVD SSML text-to-video Median Rank 6.0 #23 of 24 Archive leaderboard report
Video Retrieval MSVD SSML text-to-video R@1 20.3 #23 of 24 Archive leaderboard report
Video Retrieval MSVD SSML text-to-video R@10 63.3 #23 of 24 Archive leaderboard report
Video Retrieval MSVD SSML text-to-video R@5 49.0 #23 of 24 Archive leaderboard report
Video Retrieval MSVD SSML text-to-video R@50 -- #23 of 24 Archive leaderboard report
Visual Question Answering MSRVTT-QA SSML Accuracy 0.35 #4 of 4 Archive leaderboard report
Visual Question Answering (VQA) MSVD-QA SSML Accuracy 0.351 #33 of 36 Archive leaderboard report
Zero-Shot Video Retrieval LSMDC SSML text-to-video R@1 4.2 #16 of 16 Archive leaderboard report
Zero-Shot Video Retrieval LSMDC SSML text-to-video R@10 17.1 #16 of 16 Archive leaderboard report
Zero-Shot Video Retrieval LSMDC SSML text-to-video R@5 11.6 #16 of 16 Archive leaderboard report
Zero-Shot Video Retrieval MSR-VTT SSML text-to-video R@1 8.0 #39 of 41 Archive leaderboard report
Zero-Shot Video Retrieval MSR-VTT SSML text-to-video R@10 29.3 #39 of 41 Archive leaderboard report
Zero-Shot Video Retrieval MSR-VTT SSML text-to-video R@5 21.3 #39 of 41 Archive leaderboard report
Zero-Shot Video Retrieval MSVD SSML text-to-video R@1 13.66 #14 of 14 Archive leaderboard report
Zero-Shot Video Retrieval MSVD SSML text-to-video R@10 47.74 #14 of 14 Archive leaderboard report
Zero-Shot Video Retrieval MSVD SSML text-to-video R@5 35.7 #14 of 14 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections