In this video:
First noticing the stairway in the École polytechnique fédérale de Lausanne is surprisingly similar to the Scala logo (Martin Odersky invented Scala here), Romeo then talks with Skymind Deep Learning Engineer Francois Garillot who principally develops JVM libraries for deep learning and discusses Deeplearning4j, a library that allows you to perform deep neural network training on Apache Spark and GPUs using Java™, Scala, and Python. Francois also develops deep learning models for organizations.
Deeplearning4j is the first commercial-grade, open-source, distributed deep learning library written for Java™ and Scala. Integrated with Hadoop and Spark, it is designed to be used in business environments on distributed CPU and GPU clusters. Silicon Valley based startup Skymind is its commercial support arm.
Deeplearning4j lets you compose deep neural nets from different types of layers like restricted Boltzmann machines, autoencoders, convolutional or recurrent nets as needed in a distributed, production-grade framework that works with Spark and Hadoop on top of distributed CPUs or GPUs.
Francois explains that the higher-level deep learning framework like Keras gives users access to lower-level frameworks like TensorFlow and Theano. Deeplearning4j, the library written for Java™ and the JVM lets the programmer access those features as well from a high-level Python script, but makes use of JVM features like interoperability, functional programming, and strong static type system attributes of a language like Scala. According to Francois, with a library like Deeplearning4j:
“We are translators between lots of systems and our Swiss Army knife to do that is the JVM.”
Resources for you
- Explore Deeplearning4j, the open source, distributed, deep learning library for the JVM
- Romeo provides Deeplearning4j examples on GitHub
- Use this article as an introduction to deep learning and long-short term memory networks
- Learn about Watson machine learning
- Explore Watson machine learning on Bluemix
- Learn whether to choose deep learning and how to select the right model for your cloud
- Here’s a Python notebook that shows you how to use deep learning to build a classification model
- Try IBM Cloud for free
Follow Romeo as he tackles the most difficult challenges in data science.