MASTER
 
 

Distributed Deep Learning with deeplearning4j, 2-Day Course

By By the Bay LLC (other events)

Thu, Sep 25 2014 9:00 AM Fri, Sep 26 2014 5:00 PM
 
ABOUT ABOUT

Academy By the Bay, which runs Scala and Spark training and Scala By the Bay developer conference, is proud to present a hands-on, two-day professional training course on Distributed Deep Learning with Scala.  Our model is to partner with top experts on the best open-source implementations of algorithms immediately applicable in a wide range of startup and enterprise applications. For this course, we partner with Adam Gibson, the creator of deeplearning4j.org and founder of skymind.io.  Seating is limited to 25 participants.  We'll focus on Scala/Akka implementations and will expect substantial knowledge of software engineering and an ability to catch up on distributed systems and numeric optimization fairly rapidly.  This is the first Deep Learning course which focuses on the scale-out performance optimization made possible by the advanced distributed technologies such as Akka, the Scala-based actor programming framework.


Deeplearning4j is the first distributed, open-source deep-learning framework written for the JVM. The computational needs of deep neural networks can best be met with a massively parallel system. Deeplearning4j includes its own distributed system using Akka and AWS; similar distributed structures can be achieved with Hadoop or Spark. Deeplearning4j’s algorithms — deep-belief networks, deep autoencoders, recursive neural tensor nets and convolutional nets among others — can pretrain and/or finetune on datasets in parallel, with the entire network averaged at master. (Parameter averaging is an automatic form of regularization.) These distributed systems exhibit linear scaling; i.e. 3x workers, 3x faster learning. Next steps include integrating with multiple GPUs on AWS.

BIO

Adam Gibson is a deep-learning specialist based in San Francisco who works with Fortune 500 companies, hedge funds, PR firms and startup accelerators to create their machine-learning projects. Adam has a strong track record helping companies handle and interpret big real-time data. Adam has been a computer nerd since he was 13, and actively contributes to the open-source community through deeplearning4j.org.

This is a paid professional training course running over two full days. Food and coffee is provided, you'll need to bring your own laptop with the software preloaded -- we'll send the instructions and location closer to the date. Early Bird pricing is in effect until August 31st (+$400 from September 1st).  Except for the Academy canceling the class, the tickets are non-refundable.

The workshop is held at the Preservation Park in Oakland.  The first day is held at the Ginn House, 660 13th St.  Once you enter the preservation park from the Martin Luther King Way, the Ginn house is on the right, and the auditorium is on the ground floor.  The best way to get there is using public transportation, the nearest BART station is 12th St. City Center (http://www.bart.gov/stations/12th).