Understanding Deep Learning Phenomena

Abstract: 

Our understanding of modern neural networks lags behind their practical successes. This growing gap poses a challenge to the pace of progress in machine learning because fewer pillars of knowledge are available to designers of models and algorithms.

I work on principled approaches to theoretically and empirically investigate deep learning phenomena, using tools from ML theory and statistics. I focus on the interplay between data, training algorithms and network architecture. The goal is to improve training and generalization performance in state of the art deep learning models and extend the current success of our models to new domains. In this talk I will dive into some of my recent results in this domain and how they propose new approaches to using deep learning in practice.

Bio: 

Hanie Sedghi is a research scientist at Google Brain. She works on large-scale machine learning, especially latent variable probabilistic models. Her approach is to bond theory and practice in machine learning by designing algorithms with theoretical guarantees that also work efficiently in practice and lead the state of the art. Prior to joining Brain, she was a research scientist at Allen Institute for AI. Hanie received her PhD in Electrical Engineering from University of Southern California with a minor in Mathematics in 2015. She was also closely collaborating with Professor Anima Anandkumar at UC Irvine during her PhD studies. She received her M.Sc. and B.Sc. degrees from Sharif University of Technology, Tehran, Iran.