Y
Yue Sun
Researcher at Amazon.com
Publications - 6
Citations - 834
Yue Sun is an academic researcher from Amazon.com. The author has contributed to research in topics: Computer science & Modular design. The author has an hindex of 1, co-authored 1 publications receiving 338 citations.
Papers
More filters
Posted Content
ResNeSt: Split-Attention Networks
Hang Zhang,Chongruo Wu,Zhongyue Zhang,Yi Zhu,Zhi Zhang,Haibin Lin,Yue Sun,Tong He,Jonas Mueller,R. Manmatha,Mu Li,Alexander J. Smola +11 more
TL;DR: A simple and modular Split-Attention block that enables attention across feature-map groups ResNet-style is presented that preserves the overall ResNet structure to be used in downstream tasks straightforwardly without introducing additional computational costs.
Journal ArticleDOI
Dynamic Sparse Training via Balancing the Exploration-Exploitation Trade-off
TL;DR: In this article , the authors considered the dynamic sparse training as a sparse connectivity search problem and designed an exploitation and exploration acquisition function to escape from local optima and saddle points, and provided the theoretical guarantees for the proposed method and clarified its convergence property.
Finite Sample Identification of Low-Order LTI Systems via Nuclear Norm Regularization
Yue Sun,Samet Oymak,Maryam Fazel +2 more
TL;DR: In this paper , Hankel nuclear norm (HNN) regularization was proposed to encourage the Hankel matrix to be low-rank, which corresponds to the dynamical system being of low order.
Dynamic Sparse Training via More Exploration
TL;DR: In this paper , the dynamic sparse training is considered as a sparse connectivity search problem and an exploitation and exploration acquisition function is designed to escape from local optima and saddle points, which outperforms the SOTA sparse training methods on a wide variety of deep learning tasks.
Journal ArticleDOI
System Identification via Nuclear Norm Regularization
Yue Sun,Samet Oymak,Maryam Fazel +2 more
TL;DR: In this paper , Hankel nuclear norm regularization has been used to identify low-order linear systems via Hankel matrix regularization, which leads to new bounds on estimating the impulse response and the Hankel matrices associated with linear systems.