This page contains slides and links to videos of talks I have given.
Review lecture on Practical Problems in Deep Learning
Here is a video of a lecture I've given at the 2021 Deep Learning Theory Summer School at Princeton, where I review some important practical problems that warrant a systematic study and theoretical investigation.
Review talk on the Physics of Machine Learning
Here is a video of a talk I've given at the KIAS physics colloquium series, where I review recent advances in deep learning that benefited from methods commonly used in physics.
Talks on the Dependence of Optimal SGD Hyperparameters on Network Architecture
I've put out some work on systematically studying the dependence of optimal hyperparemters, i.e., hyperparameters (e.g., learning rate or batch size) that maximize neural network performance, on the architecture (e.g., width or depth) of the network.
Here are slides from a talk I've given at the Theoretical Physics for Machine Learning workshop at the Aspen Center for Physics.
Here is a video of a talk I've given at the Physics meets ML workshop at Microsoft Research.
Talk on Cluster Algebras and Dualities of 2D N=(2,2) Quiver Gauge Theories
Here is a video of a talk I've given at Rutgers University.
Talk on the Mordell-Weil Group of Elliptically Fibered Calabi-Yau Threefolds
These are slides for a talk I've given at the University of Pennsylvania.
Talks on The Landscape of Six-Dimensional Supergravity Theories
The main theme of my graduate research was to understand the landscape of six-dimensional supergravity theories and its relation to six-dimensional string vacua. I've collected some slides and links to videos of talks on this subject here.
This is a link to a video of a talk given at "Holographic Cosmology 2.0."
This is a link to a video of a talk given at "String Phenomenology 2012."
These are slides for a talk given at the Harvard Duality Seminar in 2012. It summarizes most of the work I've done related to this topic.
I've given talks on some unpublished results related to this topic. I've found that a topological number of a 4-manifold needs to be bounded in order for it to be used as a base of a non-singular elliptically fibered Calabi-Yau threefold. This is a video of a talk on the subject I gave. Slides of another talk I've given on this work can be found here.
For those who would like to learn more, this paper gives an overview of the program of classifying consistent six-dimensional supergravity theories and string vacua. An excellent lecture note on this subject can be found here.