Xin Wang

I am a forth year CS Ph.D. student at UC Berkeley. I am part of the RISE Lab (formerly AMP Lab), Berkeley AI Research Lab (BAIR) and Berkeley DeepDrive. I'm fortunate to be advised by Prof. Joseph E. Gonzalez and Prof.Trevor Darrell. Prior to UCB, I obtained my B.S. degree in computer science from Shanghai Jiao Tong University in 2015.

I'm broadly interested in the design of deep learning models with system motivation and its application in computer vision. My recent work focuses on model inference speedup and the design of dynamic neural networks that can adapt computation on a per-input basis. I was also involved in the Clipper project on machine learning model serving.

Office: 465 Soda Hall, Berkeley, CA 94720

Google Scholar / Github / CV


TAFE-Net: Task-Aware Feature Embeddings for Efficient Learning and Inference
Xin Wang, Fisher Yu, Ruth Wang, Trevor Darrell, Joseph E. Gonzalez
SkipNet: Learning Dynamic Routing in Convolutional Networks
Xin Wang, Fisher Yu, Zi-Yi Dou, Trevor Darrell, Joseph E. Gonzalez
European Conference on Computer Vision (ECCV) 2018
IDK Cascades: Fast Deep Learning by Learning not to Overthink
Xin Wang, Yujia Luo, Daniel Crankshaw, Alexey Tumanov, Fisher Yu, Joseph E. Gonzalez
Conference on Uncertainty in Artificial Intelligence (UAI) 2018
Clipper: A Low-Latency Online Prediction Serving System
Daniel Crankshaw, Xin Wang, Guilio Zhou, Michael J. Franklin, Joseph E. Gonzalez, Ion Stoica
USENIX Symposium on Networked Systems Design and Implementation (NSDI) 2017
Scalable Training and Serving of Personalized Models
Daniel Crankshaw, Xin Wang, Joseph E. Gonzalez, Michael J. Franklin
LearningSys 2015