Matt JonesSoftware Engineer at Pinterest
Matt is a software engineer at Pinterest, where he builds systems that ensure pinners experience a safe and spam-free site. He's interested in building data-driven systems, thinking about distributed computing, and finding novel applications for sous-vide cooking. Before starting at Pinterest, Matt hacked on search and mobile stuff at Yelp.
Machine Learning for Humans
Location:Bayview BDuration:Half DayAbstract:Machine learning makes human decisions and insights scalable. Simple applications of a few basic techniques can dramatically improve productivity and simplify rote tasks throughout your organization. This tutorial will focus on Vowpal Wabbit, a widely-used classification and regression package. You'll learn how to identify problems where basic ml tools can automate repetitive tasks, what kind of data you'll need to ensure success, and how to evaluate and report on performance. You'll also learn how to integrate insights from across your organization into machine-learning systems, and how tools like Vowpal Wabbit make solving fuzzy problems fun. The tutorial will start with a walkthrough of how we aggregate and use data in the war on abusive content at Pinterest, and an overview of some of the systems we've built along the way. You'll then learn how Vowpal Wabbit implements machine learning techniques, and how to map your problems to regression and classification problems. You'll use VW to train a model on a sample dataset and learn how to evaluate and optimize the performance of this model. Finally, you'll learn how to monitor the models you've built, and how to integrate these models into production systems.