Optum/UHG:

Graph Day and Workshops

September 27, 2018

Note: Separate registration is required for Graph Day and workshops.

Graph Day (All Day)

Databricks & Microsoft Workshop (Afternoon)

1:00-5:00pm
Rooms 1&2
13625 Technology Drive, Eden Prairie, MN 55344 (MN051 Building)

Introduction to Azure Databricks with Data Science
With the rise of large biomedical datasets, many researchers are turning to cloud computing and scalable computing technologies like Apache Spark. In this talk, we will introduce the Azure Databricks Unified Analytics Platform, which simplifies the process of running Apache Spark on the Microsoft Azure cloud. We will then demonstrate how Apache Spark can be used to analyze a large corpus of data. By using Apache Spark, we can rapidly cluster patients based on lab values, providing insight into disease severity, progression, and provider efficacy.

Lead Presenter: Frank Nothaft
Assisting Presenter: Wesley Dias

Detecting Melanoma with Transfer Deep Learning
This presentation will showcase a bootstrapped TensorFlow pre-trained model (InceptionV3) being used to classify between three classes of images (melanoma, seborrheic, and normal) in order to aid in diagnosis of suspect images of abnormal skin conditions that could indicate carcinoma.

Key technologies employed: Spark Pipelines,Spark Deep Learning, Pandas, and SparkML Random Forest.
Transfer Deep Learning Demo

Lead Presenter: Ben Wilson, Yongsheng Huang
Assisting Presenters: Frank Nothaft, Christian Nimsch

Analyzing Medical Abstracts with NLP + Spark
This code demonstration shows how to ingest, parse, tokenize, vectorize, and apply an LDA model to medical texts, extracting important keywords to summarize the critical topics contained within corpuses of medical journals.

Key technologies employed: Spark NLP, LDA Models, and Spark Feature Engineering.
Medical NLP Demo

Lead Presenter: Ben Wilson, Yongsheng Huang
Assisting Presenters: Christian Nimsch, Wesley Dias

End-to-end analysis of genomic data using the Databricks UAP4Genomics
Genome sequencing data can provide an unprecedented lens into an individual’s disease risks, but the end-to-end analysis of large scale genomics data remains complex and expensive. In this workshop, we will walkthrough how the Databricks Unified Analytics Platform for Genomics (UAP4Genomics) simplifies the end-to-end process of turning raw sequencing data into actionable insights at population scale. We will demonstrate how to call variants in a single sample using our accelerated GATK4 pipeline, before using tools like Hail to characterize the association of variants in a population with clinical phenotypes. Once we identify variants that associate with phenotypes that we are interested in, we will then use machine learning to model genome-wide disease risk in a reproducible manner.

Key technologies employed: GATK4/Variant calling, Hail/Genotype-phenotype association tests, population scale risk-modeling via ML, ML model training/deployment

Lead Presenter: Frank Nothaft
Assisting Presenters: Karen Feng

Informatica Workshop (Afternoon)

Intel / Swarm64 Workshop (Morning)

Rooms 1&2
13625 Technology Drive, Eden Prairie, MN 55344 (MN051 Building)

Intel AI Update 8:30-9:15am
Want to accelerate AI workloads 10, 100 or 200 times faster on a ubiquitous Intel Xeon processor? We’ve got you covered. Find out the latest tools, techniques, resources, and roadmaps across Intel’s entire AI roadmap.

Speakers: Andy Bartley – Healthcare solution architect and Greg Reiff – Solution Engineer

Intel – Go faster with with FPGAs (101 Class) 9:15-10:00am
CPUs aren’t fast enough! What else do you have Intel? You’re in luck. Intel is meeting that challenge head on is with the use of FPGA or Field Programmable Gate Arrays. This technology allows YOU to create your own custom/dedicated hardware or processing functions from the edge to the cloud! “But I’m not an FPGA programmer???!!!” We’ve got you covered.

Through complete software infrastructure for deployment into the datacenter to high level design flows with AI frameworks like Caffe, Tensorflow and C++ and OpenCL flows – you know longer need to be an FPGA engineer to harness the power of FPGA based acceleration. Come to this introductory session and to see the world of possibilities and then stay for a deeper dive into OpenCL based acceleration.

Speaker: Jeff Nigh, FPGA Account Manager

Technical Deep dive – OpenCL Acceleration with FPGAs 10:00-10:45am
Scott, our local Minnesota guru in the Programmable Solutions Group will walk you through tips and tricks for OpenCL based acceleration on FPGAs. To truly harness FPGA acceleration with OpenCL you need to be armed with a little bit of knowledge to set up structures to release the parallel compute power available in the FPGA.

Speaker: Scott Prigmore, FPGA Guru

Database/Analytics Acceleration with FPGAs 10:45-11:30am
I need my database and analytics to go faster! HELP! This session will introduce you to solutions from Intel’s partners for enterprise based acceleration such as MySQL, PostgreSQL, MariaDB, Cassandra, Elasticsearch along with Spark and Hadoop acceleration. These partners solutions have done all the work creating custom acceleration functions for the FPGA along with the software to very seamlessly integrate into your current enterprise/analytics workflows. Drop in an FPGA and go!

Speaker: Jeff Nigh, FPGA Account Manager

TensorFlow, Sessionization and Npath – Advanced Analytics Hands On Workshop (All Day)

UDW Workshop (Morning & Afternoon)

Prior Authorizations and Machine Learning in the Clinical Space (Afternoon)

If the workshop you would like to attend is sold out, you can join the waiting list here.

Presented by Optum as conceived by MinneAnalytics.

Optum/UHG:

Graph Day and Workshops

September 27, 2018

Room 8

Room 7