12 events found.
A Mathematical Exploration of Why Language Models Help Solve Downstream Tasks
https://youtu.be/OoimTbnSe7I Speaker: Nikunj Saunshi, Dept. of Computer Science, Princeton University Title: A Mathematical Exploration of Why Language Models Help Solve Downstream Tasks Abstract: Autoregressive language models pretrained on large corpora have been successful at solving downstream tasks, even with zero-shot usage. However, there is little theoretical justification for their success. This paper considers the following […]
2/16/2021 Computer Science for Mathematicians
Speaker: Michael P. Kim (UC Berkeley) Title: Outcome Indistinguishability Abstract: Prediction algorithms assign numbers to individuals that are popularly understood as individual “probabilities” — e.g., what is the probability of 5-year survival after cancer diagnosis? — and which increasingly form the basis for life-altering decisions. The understanding of individual probabilities in the context of such unrepeatable events […]
Global Anomalies on the Hilbert Space
Speaker: Jaume Gomis (Perimeter PI) Title: Global Anomalies on the Hilbert Space Abstract: We will discuss an elementary way of detecting some global anomalies from the way the symmetry algebra is realized on the torus Hilbert space of the anomalous theory, give a physical description of the imprint of the “layers” that enter in the […]