Timezone: »

 
Poster
Matryoshka Representation Learning
Aditya Kusupati · Gantavya Bhatt · Aniket Rege · Matthew Wallingford · Aditya Sinha · Vivek Ramanujan · William Howard-Snyder · Kaifeng Chen · Sham Kakade · Prateek Jain · Ali Farhadi

Tue Nov 29 09:00 AM -- 11:00 AM (PST) @ Hall J #640
Learned representations are a central component in modern ML systems, serving a multitude of downstream tasks. When training such representations, it is often the case that computational and statistical constraints for each downstream task are unknown. In this context rigid, fixed capacity representations can be either over or under-accommodating to the task at hand. This leads us to ask: can we design a flexible representation that can adapt to multiple downstream tasks with varying computational resources? Our main contribution is Matryoshka Representation Learning (MRL) which encodes information at different granularities and allows a single embedding to adapt to the computational constraints of downstream tasks. MRL minimally modifies existing representation learning pipelines and imposes no additional cost during inference and deployment. MRL learns coarse-to-fine representations that are at least as accurate and rich as independently trained low-dimensional representations. The flexibility within the learned Matryoshka Representations offer: (a) up to $\mathbf{14}\times$ smaller embedding size for ImageNet-1K classification at the same level of accuracy; (b) up to $\mathbf{14}\times$ real-world speed-ups for large-scale retrieval on ImageNet-1K and 4K; and (c) up to $\mathbf{2}\%$ accuracy improvements for long-tail few-shot classification, all while being as robust as the original representations. Finally, we show that MRL extends seamlessly to web-scale datasets (ImageNet, JFT) across various modalities -- vision (ViT, ResNet), vision + language (ALIGN) and language (BERT). MRL code and pretrained models are open-sourced at https://github.com/RAIVNLab/MRL.

Author Information

Aditya Kusupati (University of Washington Google Research)
Gantavya Bhatt (University of Washington)

Graduate Student (currently working from home) in ML, Submodularity and NLP

Aniket Rege (University of Washington)
Aniket Rege

Graduate researcher in the RAIVN Lab at the University of Washington working on representation learning for the visual world.

Matthew Wallingford (University of Washington)
Aditya Sinha (University of Illinois Urbana Champaign)

Research Fellow at Google Research India, interested in Machine Learning and Optimisation

Vivek Ramanujan (Department of Computer Science, University of Washington)
William Howard-Snyder (Department of Computer Science, University of Washington)
Kaifeng Chen (Google)
Sham Kakade (Harvard University & Microsoft Research)
Prateek Jain (Google Research)
Ali Farhadi (University of Washington, Allen Institute for Artificial Intelligence)

More from the Same Authors