Self-supervised Learning from a Multi-view Perspective

By Yao-Hung Hubert Tsai, Yue Wu, Ruslan Salakhutdinov, Louis-philippe Morency. 2020

234 citations recorded by the source corpus.

Subfields: Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Representation & Self-Supervised Learning

Citations in the corpus

HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units (2021, background): “These objectives can usually be interpreted as mutual information maximization [52].”
What Makes Multimodal Learning Better than Single (Provably) (2021, background): “Moreover, the main assumption behind previous analysis [37, 39, 42] is that each view/modality contains sufficient information for target tasks, which does not shed light on our analysis.”
A Survey on Self-Supervised Learning: Algorithms, Applications, and Future Trends (2023, unknown): “Several studies have delved into this area [54], [71], [72], [73], [74], [75], [76], [77], [78], [79], [80], [81].”
Conditional Contrastive Learning: Removing Undesirable Information in Self-Supervised Representations (2021, background): “Examples of the unrelated data include different images [5, 18] or an image and a caption from another image [43].”
Hubert: How Much Can a Bad Teacher Benefit ASR Pre-Training? (2021, background): “These objective can usually be interpreted as mutual information maximization [37].”
A Survey on Contrastive Self-supervised Learning (2020, background): “[84] provide an information-theoretical framework from a multi-view perspective to understand the properties that encourage successful self-supervised learning.”
COMPLETER: Incomplete Multi-view Clustering via Contrastive Prediction (2021, methodology): “As one of most effective unsupervised learning paradigms, contrastive learning [2, 4, 8, 23, 28, 30, 37, 38] has achieved state-of-the-art performance in representation learning.”
A Brief Overview of Unsupervised Neural Speech Representation Learning (2022, background): “Self-supervised learning is a subset of unsupervised learning (Tsai et al. 2020).”
Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style (2021, background): “1), our problem setting is more general: [3, 74] assume (approximate) independence of views (x, x̃) given the task (content), while [118, 119] assume (approximate) independence between one view and the task (content) given the other view, neither of which hold in our setting.”
From Canonical Correlation Analysis to Self-supervised Graph Neural Networks (2021, background): “(5) with the multi-view Information Bottleneck [9] and the minimal and sufficient representations for self-supervision [45]: Corollary 1.”
Lossy Compression for Lossless Prediction (2021, background): “This is related to the multi-view literature, where one only needs to retain information which is invariant across views [30, 31, 118, 119].”
Dual Contrastive Prediction for Incomplete Multi-View Representation Learning (2022, background): “On the other hand, our theoretical result is also different from [36].”
High-Dimension Human Value Representation in Large Language Models (2024, methodology): “Following the information bottleneck principle for representation learning [71, 72], we consider a surrogate task named value embedding learning [73].”