IFIP ICCIDS 2021

PRE-CONFERENCE WORKSHOP ON

How Deep Learning helps finding patterns in Time-Series data? – Exploring with Sparkling-Water (H2o.ai)

Abstract:
While processing and analyzing non-linear time-series data, finding interesting patterns by using deep neural networks is still an interesting problem to solve. Many of the today’s big data storage systems like HDFS, S3, GCS or ADLS are distributed, or the object storage is spread across a cluster. Training a deep-learning model with ReLU or Sigmoid function on the big data spread across a cluster necessitates distributed execution of Deep Learning algorithms unlike running on GPUs. H2o.ai introduces a new framework called sparkling-water which is built on top of Apache Spark, an open-source distributed data processing framework and h20.ai flow that provides deep learning algorithms. This workshop aims to showcase an Airline use case of predicting the arrival delay using Sparkling-water.