LA Machine Learning Meet-up

FIELD REPORT

I had an excellent alternative to all the media machinations on Election Day Nov. 6 2012; I went to a machine learning event! Hosted by the LA Machine Learning group LA Machine Learning group  on Meetup.com, the event went by the name of “The Unreasonable Effectiveness of Ensembles.” This topic was of specific interest to me because many of the entries on the Kaggle leaderboard for the $3 million Heritage Health data science competition use ensembles. I wanted to refine my knowledge about this useful ML technique.

The event took place at the Century City offices of Factual.com, a company that provides access to data for powering web and mobile apps, mobile advertising, and enterprise solutions. I arrived early to get a jump on the vibe of this Meetup group since this was my first event. The Factual office was perfect for this kind of thing, complete with a large meeting area that had an ample lecture space, bean bag chairs, piano and other games designed to please the developer geeks working there. Factual did it right by providing plenty of pizza and beer. One guy had his tablet tuned to Nate Silver’s blog to monitor the election results.  After chatting with a number of fellow data scientists, the talk was about to begin.

The lecturer was Rudiger Lippert, a software developer at Factual.  Rudiger studied Electrical Engineering at Boston University and went on to get a Master’s degree at UCLA, specializing in Signal Processing. In graduate school his research centered around Speech Recognition.  Rudi’s talk was excellent and covered all the areas of ensembles I had hoped for.

Ensemble methods are considered by many to be the most important development in machine learning of the last decade. By combining many weak models to produce a single strong model, ensemble methods have performance which rivals and very often beats that of other model classes such as Support Vector Machines (SVM) and Neural Networks. The talk started with simple Decision Trees, and went on to Bagging, Random Forests, Boosting, and newer developments such as Regularized Greedy Forests. It was a great overview of the subject and one that I plan to utilize in my consulting practice.

I would definitely recommend this Meetup group, but if you’re not from LA, try to find a similar group in your area.

Advertisements

Posted on November 20, 2012, in Field Report, Uncategorized and tagged . Bookmark the permalink. Leave a comment.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Starbridge Partners

Welcome to STARBRIDGE PARTNERS! We Help Build Careers in Data Science. We Specialize in Placing Data Scientists & Data Engineers.

Follow the Data

A data driven blog

p-value.info

Sponsored by AMULET Analytics

Simply Statistics

Sponsored by AMULET Analytics

Data Science 101

Learning To Be A Data Scientist

blog.RDataMining.com

R and Data Mining

Sponsored by AMULET Analytics

%d bloggers like this: