Welcome

by Séb Arnold, October 13, 2016

I am an undergraduate student at the University of Southern California, double majoring in Computer Science and Math. My research interests lie in artificial intelligence, distributed systems, and mathematical optimization.

I also like skiing (a lot).

[Contact, Resume]

Me and motorcycle hair

The all-new motorcycle haircut




News

ICLR17 Workshop submission

Together with Prof. Chunming Wang, we submitted a workshop paper presenting our work on second-order optimization for distributed deep learning to ICLR17. Update: It was accepted! [Link]


SoCal Robotics Symposium

Together with Ewina Pun and Théo Denisart, we will be presenting Shapechanger - our Environments for Transfer Learning - at the SoCal Robotics Symposium.


Open-Sourcing mj_transfer

We open-sourced some MuJoCo environments for robotic transfer learning on OpenAI Gym. [GitHub Repo, Website]


Open-Sourcing randopt

I open-sourced randopt, a python package to streamline the search for good hyper-parameters. An effort to be continued.




Projects

randopt

Randopt is a python package to streamline the search for good hyper-parameters. It provides a programmatic interface to serialize, choose, and visualize hyper-parameters and results. [GitHub Repo, Blog Post]

Paperify.py

A tool to convert markdown files into beautiful LaTeX docs and web pages. A work in progress. [Web Demo, PDF Examples]

Specialists

Code for the report "A Greedy Algorithm to Cluster Specialists", where I develop an algorithm for the class assignment problem in the generalist-specialist framework. [GitHub Repo]

Quantum Opt

Comparing the performance of simulated annealing against a simulation of adiabatic quantum computing on Ising problems. [GitHub Repo]

Tooski

Simply the best skiing website ever ? [Website]

ML Compare

Compare machine learning algorithms on the same dataset, with only one line. It is distributed across your cores ! [GitHub Repo]

Meritocracy

A web-based social experiment to evaluate the meritocracy of people. [GitHub Repo]

Generatic Experiment

Uses a genetic algorithm to generate images out of a trained convnet. Also, tries to learn from the generated images only. [GitHub Repo]

Tea Stock Market

Predict the fluctuations of Indian Tea Companies' stock market, based on weather data. A lot of machine learning, some financial analysis. [GitHub Repo]