• Yong Zhuang: Data science is not a panacea for all problems in science and engineering, but used in the right way, it provides a principled approach to maximally leverage the data that we have and inform what new data to collect.source
• Tianyu Kang: This analysis generates curves of accuracy as a function of the number of directions ablated: the more reliant a network is on low-dimensional activation subspaces, the more quickly the accuracy will drop as single directions are ablated..source
• Chengjie Zheng: Through extensive experiments, we demonstrate that using an explicit few-shot learner and action appearance aligned features makes MetaUVFS significantly better suited for downstream few-shot tasks compared to all state-of-the-art unsupervised methods.source
• Wei Ding: The fundamental purpose of scientific discourse is not the mere presentation of information and thought, but rather its actual communication. It does not matter how pleased an author might be to have converted all the right data into sentences and paragraphs; it matters only whether a large majority of the reading audience accurately perceives what the author had in mind. Therefore, in order to understand how best to improve writing, we would do well to understand better how readers go about reading. Such an understanding has recently become available through work done in the fields of rhetoric, linguistics and cognitive psychology. It has helped to produce a methodology based on the concept of reader expectations.source