I am Haozhe Jiang, a second-year EECS PhD student at UCBerkeley. I want to understand why deep learning works in principle. I try to develop theory that has implications to real-world models and discover empirical laws that extrapolate well. I am also enthusiastic about Physics and work on applying machine learning to other fields such as cryptography and astronomy.

Please feel free to reach out to me.

Biography

I was born and raised in the wonderful city of Shanghai. My childhood dream was to become an astronomer and it triggered my interest in Science when I was still a little boy. Before college, I was at the No.2 High School of East China Normal University. Back then, I took part in Physics Competition and I was deeply impressed by the beauty of Physics. In my freshman and sophomore years, I worked primarily with Professor Chongjie Zhang on empirical Reinforcement Learning. Later I worked closely with Professor Simon Du and Professor Maryam Fazel on learning problems in Game Theory. One of the most exciting projects I have ever done is breaking a fundamental cryptographic hard problem, the Learning Parity with Noise Problem with Neural Networks, in the high-noise regime. It was initially a course project of Fundamentals of Cryptography instructed by Professor Yilei Chen. By the way, his homepage is very fun to read and I highly recommend you check it out. In my undergrad thesis I tried to use machine learning to find anomaly microlensing events advised by Professor Shude Mao. I feel like chasing my dream as a little boy again.

Now I am coadvised by Nika Haghtalab and Jiantao Jiao at Berkeley. More recently I am also a visiting researcher at FAIR.

Miscellaneous

My favorite sport is Basketball. I usually play as a shooting guard.

I love classical music. My favorite piece is Op.53 by Frédéric Chopin, Polonaise in A Flat. Recently I am also brainwashed by Op.44, another fantastic Polonaise. My favorite composer is Ludwig van Beethoven.

Photographying is one of my hobbies. I mostly shoot natural views. Check out some here.