Researchers from the Machine Learning Center at Georgia Tech (ML@GT) will present seven papers at the Ninth Annual International Conference on Learning Representations (ICLR). Accepted research will present new findings on topics such as increasing multi-platform deployment and semi-supervised object detection.
[RELATED STORY: New Model Defense Strategy Prevents Cloning]
ML@GT Associate Director and School of Interactive Computing (IC) Assistant Professor Zsolt Kira and Tuo Zhou, an assistant professor in the H. Milton Stewart School of Industrial Systems and Engineering will serve as conference area chairs. IC Associate Professor Devi Parikh and School of Computational Science and Engineering Associate Professor Polo Chau are steering committee members of the Rethinking ML Papers workshop.
The workshop will focus on answering questions like how to design an inclusive publication format for ML research, how to communicate ML research and theory more effectively, and how best to interpret complex information in a visual setting.
“ML@GT continues to publish cutting-edge research across many fields. We are looking forward to sharing this recent work at a venue as well regarded as ICLR,” said Kira.
Accepted papers from Georgia Tech are:
- CompOFA – Compound Once-For-All Networks for Faster Multi-Platform Deployment by Manas Sahni, Shreya Varshini, Alind Khare, and Alexey Tumanov (all Georgia Tech)
- Protecting DNNs from Theft using an Ensemble of Diverse Models by Sanjay Kariyappa (Georgia Tech), Atul Prakash (University of Michigan), and Moinuddin K Qureshi (Georgia Tech).
- Creative Sketch Generation by Songwei Ge (University of Maryland, College Park), Vedanuj Goswami (Facebook AI Research), C. Lawrence Zitnick (Facebook AI Research), and Devi Parikh (Georgia Tech and Facebook AI Research)
- Large Batch Simulation for Deep Reinforcement Learning by Brennan Shacklett (Stanford University), Erik Wijmans (Georgia Tech), Aleksei Petrenko (University of Southern California), Manolis Savva (Simon Fraser University), Dhruv Batra (Georgia Tech), Vladlen Koltun (Intel Labs), Kayvon Fatahalian (Stanford University)
- Unbiased Teacher for Semi-Supervised Object Detection Yen-Cheng Liu (Georgia Tech), Chih-Yao Ma (Facebook), Zijian He (Facebook), Chia-Wen Kuo (Georgia Tech), Kan Chen (Facebook), Peizhao Zhang (Facebook), Bichen Wu (Facebook), Zsolt Kira (Georgia Tech), Peter Vajda (Facebook)
- Molecule Optimization by Explainable Evolution by Binghong Chen (Georgia Tech), Tianzhe Wang (Georgia Tech), Chengtao Li (Galixir), Hanjun Dai (Google Brain), Le Song (Georgia Tech)
- A Hypergradient Approach to Robust Regression without Correspondence by Yujia Xie (Georgia Tech), Yixiu Mao (Georgia Tech), Simiao Zuo (Georgia Tech), Hongteng Xu (Georgia Tech), Xiaojing Ye (Georgia Tech), Tuo Zhao (Georgia Tech), and Hongyuan Zha (The Chinese University of Hong Kong, Shenzhen)
[RELATED: Watch short summaries of accepted papers]
ICLR is dedicated to advancing deep learning and publishing cutting-edge research that impacts data science, artificial intelligence, and statistics. The conference will be held virtually May 3-7.
Story by Allie McFadden, Communications Officer, email@example.com