SCOPE
Dr. Teney’ talk will demonstrate how tasks such as image captioning and visual question answering can serve as testbeds for complex reasoning, out-of-distribution generalization, and robustness to distribution shifts. He will present various approaches to expand the limits of statistical machine learning, such as causal reasoning and meta learning. The aim of his work is to go beyond the classical i.i.d. setting, learn models that generalize out-of-distribution, and increase their robustness to the distribution shifts encountered in the real world.