30. Apache Beam (unified Batch and Stream processing!)

Agenda

• 17.45: Drink, socialize

• 18.00: First talk: Apache Beam (unified Batch and Stream processing!)

Abstract: Apache Beam (unified Batch and Stream processing!) is a new Apache incubator project. Originally based on years of experience developing Big Data infrastructure within Google (such as MapReduce, FlumeJava, and MillWheel), it has now been donated to the OSS community at large.

Learn about the fundamentals of out-of-order stream processing, and how Beam’s powerful tools for reasoning about time greatly simplify this complex task. Beam provides a model that allows developers to focus on the four important questions that must be answered by any stream processing pipeline.

Speaker: Daniel is a Developer Advocate at Google. With more than fifteen years of experience in the software industry, Daniel has held positions at companies such as Ericsson and Opera Software. Daniel holds a Bachelor's degree in Computer Science from Uppsala University.

• 18.45: Eat, drink, socialize (more)

• 19.00: Second talk: ID2223: A report from the world's first university course (at KTH) with practical work on Deep Learning and Big Data

Abstract: This session will introduce some of the work that was done in a course at KTH on Deep Learning and Big Data: ID2223 (Large Scale Machine Learning and Deep Learning). ID2223 was Sweden's first course on Deep Learning and, to the best of our knowledge, the only course in the world that combines practical experience of Big Data and Deep Learning. 130 students took the course and carried out a wide variety of projects, combining deep learning and Big Data. Kim and Konstantin, who won the best project award for 2018, will present their project on predicting human actiivity on edge devices using LSTMs.

Robin Andersson will also introduce the platform, Hops Hadoop, where students ran TensorFlow experiments on GPUs. The cluster is hosted at www.hops.site by SICS North in Luleå and available for use by researchers and companies in Sweden. Hops is the world's fastest Hadoop platform and only one to support GPUs-as-a-Resource.

Speakers:
KTH: Jim Dowling, Konstantin Sozinov, Kim Hammer

Dr Jim Dowling will introduce some of the projects ID2223 (Large Scale Machine Learning and Deep Learning), Sweden's first course on Deep Learning and, to the best of our knowledge, the only course in the world that combines practical experience of Big Data and Deep Learning.
Logical Clocks: Robin Andersson

Robin led the development of support for GPUs as a schedulable and isolable resource in Hops Hadoop, making it the first Hadoop distribution to natively support GPUs-as-a-resource. He has also been working on providing TensorFlow-as-a-service on the Hops platform, including Deep Learning frameworks such as Yahoo's TensorFlowOnSpark and Uber's Horovod.

• 19.45: Drink, socialize (even more)