## Overview

I am a PhD student at the University of Washington, working with Carlos Guestrin on machine learning problems. I plan to graduate at the end of summer.

I research optimization algorithms for scalable machine learning. My work focuses on practical but principled ways to train models faster by exploiting the problem's structure.

For my most recent project, I accelerated the training of deep models by adaptively prioritizing challenging training instances. Before that, I designed a fast and elegant algorithm for exploiting the sparsity of solutions in sparse optimization problems. You may like using my BlitzML solver to solve your Lasso-type and linear SVM problems.

During graduate school, I completed some interesting internships. At Amazon, I prototyped an important tool for broad use within the company. The project moved into production after the internship. At GM, I developed some ideas for autonomous vehicles near my hometown in Michigan.

Before graduate school, I studied engineering at the University of Michigan. There I worked on some research projects with Clayton Scott.

## Publications

Training Deep Models Faster with Robust, Approximate Importance Sampling.

In preparation.

A Fast, Principled Working Set Algorithm for Exploiting Piecewise Linear Structure in Convex Problems.

Preprint.

StingyCD: Safely Avoiding Wasteful Updates in Coordinate Descent.

International Conference on Machine Learning, 2017.

author = {Tyler B. Johnson and Carlos Guestrin},

title = {StingyCD: Safely Avoiding Wasteful Updates in Coordinate Descent},

booktitle = {International Conference on Machine Learning},

year = {2017}

}

Unified Methods for Exploiting Piecewise Linear Structure in Convex Optimization.

Advances in Neural Information Processing Systems 29, 2016.

author = {Tyler B. Johnson and Carlos Guestrin},

title = {Unified Methods for Exploiting Piecewise Linear Structure in Convex Optimization},

booktitle = {Advances in Neural Information Processing Systems 29},

year = {2016}

}

Blitz: A Principled Meta-Algorithm for Scaling Sparse Optimization.

International Conference on Machine Learning, 2015.

author = {Tyler B. Johnson and Carlos Guestrin},

title = {Blitz: A Principled Meta-Algorithm for Scaling Sparse Optimization},

booktitle = {International Conference on Machine Learning},

year = {2015}

}

## Software

Use my BlitzML package to quickly solve Lasso, sparse logistic regression, and linear SVM problems.

## Teaching

I have enjoyed helping teach the following courses:

- EE 578: Convex Optimization, University of Washington, Winter 2014
- CSE 546: Machine Learning, University of Washington, Fall 2013
- ENGR 100: Music Signal Processing, University of Michigan, Fall 2010
- EECS 314: Electronic Circuits, University of Michigan, Winter 2010