Distributed Deep Learning with Keras/TensorFlow on Spark: yes you can!
Technical talk | English
Technical talk | English
Theatre 16: Track 4
Thursday - 13.55 to 14.35 - Technical
Healthcare
Python is the most popular programming language of choice among data scientists. It is easy to learn, it allows fast model prototyping and several libraries for Deep Learning are available for it. But when it comes to move things to production, in particular when you need to train deep neural networks in a distributed fashion through Apache Spark, well, that’s another story. In this talk I will walk through the process of using DeepLearning4J to import Python (Keras and TensorFlow) models and successfully train them in Spark without changes to the original code.