Generalization in Deep Learning
Understanding the generalization of deep learning is an important but challenging problem. We establish generalization error bounds for SGD by characterizing the algorithmic stability in terms of the population risk at initialization and from information…
Game Testing
As modern games keep growing both in size and complexity, it has become more challenging to ensure that all the game content is well tested. To this end, we explore to develop new and efficient…
Fast Pretraining
Unsupervised language pre-training has been widely adopted by many machine learning applications. However, as the pre-training task requires no human labeling effort, a massive scale of training corpus from the Web can be used to…
Drug Discovery
Drug discovery is an important area with big business and social impact. From designing a new drug until it is finally approved, it takes tens of years and billions of dollars. We would like to…
Crystal Structure Design
Materials play an important role in energy storage and carbon capture. Particularly, energy storage is indispensable for efficient use of renewable energy and finding better materials for energy storage has long been hot research topics.…
Causal Learning-Inference
Current AI models can learn spurious correlation during data-fitting process, which can incur the failure of generalizing to new unseen domains. To resolve this problem, we resort to causal inference, with the expectation to learning…
Carbon Neutrality
As Paris Agreement entered into force, countries take further steps to limit Greenhouse Gases emissions. China pledges to peak its CO2 emissions by 2030 and achieves net-zero no later than 2060, implying rapid and dramatic…
Bio Embedding
Life is ruled by biological sequences and molecules, i.e. DNA, RNA, and protein sequences, following the de facto ‘natural’ language of biology. Understanding how these biomolecular behaves and interacts with each other can help with…
AutoML-NAS
AutoML, which aims to automate a machine learning system, has attracted a lot of attention in the research community and made a lot of noise in industry and media. An ML system typically learns from…