Adam Gibson, DeepLearning4j on Spark and Data Science on JVM with nd4j, SF Spark @Galvanize 20150212

-----

DeepLearning4j is an open source deep learning platform for the JVM focusing on data at scale and ease of use and maintainability. In this talk we will be covering the core architecture of deeplearning4j on Spark. This will cover the 3 core components of deeplearning4j:

• The core algorithms in deeplearning4j

• Distributed Deep Learning, using Jeff Dean Style Parameter averaging

Each part of the talk will involve brief live coding to demonstrate concepts and takeaways. We will end with a runtime comparison of MLlib and deeplearning4j on Spark.