Presented by DAMA PDX & NASA scientists

Sponsored by Snowflake

Calling all data enthusiasts and/or space nerds in Portland, Oregon and beyond. Join us for an exciting, all-day virtual event with NASA scientists. Attendees will work with actual NASA data and have an opportunity to collaborate on matters of data analysis directly with our presenters.

NOTE: Are you a student or between jobs? If so, you can attend for free. Please email president@damapdx.org for the registration code.

Opening Remarks, 8:30 AM by @DAMAPDX Board

Morning Session: Presentation & Workshop, 9:00 AM - 12:00 PM with Eric Lyness and Victoria Da-Poian, NASA

They will present their talk, Machine Learning to Find Life on Mars and Beyond. In 2021 the European ExoMars rover will land on Mars with the Mars Organic Molecular Analyzer (MOMA) laboratory to analyze soil samples searching for past or present life. NASA is developing machine learning algorithms to help the scientists more quickly analyze the data when it arrives from Mars. In this talk, we will present the current work using MOMA mass spectrometer data acquired during ground testing. Using this data we are aiding the scientists by matching new spectra with the most similar spectra from past experiments. We will present the nuances of mass spectra, our limitations with respect to data, and our approach to the problem. We hope to elicit feedback from the attendees.

More details at https://phys.org/news/2020-06-nasa-life-mars.html

[60 minute Break]

Sponsor Session: Snowflake Presentation, 1:00-1:30 PM with Drew Swanson and Brian Whittington, Snowflake

Afternoon Session: Presentation followed by Q&A, 1:30 PM - 3:30PM

Machine Learning at Goddard: an Approach for Detecting Wildfires with CubeSats, Enabling Science Autonomy Onboard Communication Limited Spacecraft, and Measuring Worldwide Vegetation Structure

A conversation with James MacKinnon.

Session will run from 1:30 -3:30 PT - last 30 minutes for Q&A

Abstract: This talk will primarily focus on the challenges of using machine learning techniques for the detection of wildfires from space. Wildfires are destructive to both life and property, which necessitates an approach to quickly, and autonomously detect these events from orbital observatories. This talk will introduce a neural network based approach for classifying wildfires in MODIS multispectral data, and show how it could be applied to a constellation of low-cost CubeSats. The approach combines training a deep neural network on the ground using high performance GPUs, and deploying the trained model to a highly optimized inference system running on a flight-proven embedded system. Normally neural networks execute on hardware orders of magnitude more powerful than anything found in a space-based computer, therefore the inference system is designed to be performant even on the most modest of platforms. This implementation is able to be significantly more accurate than previous neural network implementations of wildfire detection, while also approaching the accuracy of the state-of-the-art MODFIRE data products. In addition to this, the talk will also give an overview of other AI/ML work being performed at several projects across Goddard, including applying machine learning to mass spectrometer data, hyperspectral imagery, and lidar point cloud data.

Closing Remarks and Next Steps: 3:45 - 4:00 PM by DAMA PDX Board

Times above may be revised in the days before our event.

Event will be presented via Zoom. Analytical discussions and collaboration will be hosted via a Slack workspace that will begin on event day and last for a few weeks afterwards. This is a BYOAE event (bring your own analytical environment). However, attendees will have access to the data via Snowflake and to notebook templates via Zepl.

Pre-event tasks for analysis setup will be shared on 10/21 via email.

Speakers

Eric Lyness is the NASA software and operations lead at Goddard Space Flight Center for the Mars Organic Molecule Analyzer (MOMA) on the ExoMars 2020 rover.

Victoria Da-Poian is an Aerospace Engineer with NASA developing machine learning and data analysis tools for mass spectrometry data from Exomars MOMA mass spectrometer instrument.

James MacKinnon is a NASA computer engineer working in the Science Data Processing Branch at the Goddard Space Flight Center, specializing in building on-board processing systems and is focusing on both space-qualified software and hardware development. Recently he has been applying artificial intelligence and machine learning techniques to a variety of in-situ science data processing problems. He has worked on many research projects including wildfire detection, semantic segmentation of hyperspectral imagery, time series lidar analysis, and "quick look" fast science product generation. He received both his B.S and M.S at the University of Florida in Gainesville, where he helped develop several ISS technology demonstrator payloads while working for the Center for High Performance Reconfigurable Computing.

Where

Virtual event, RSVP for Zoom registration details

When

Date – Thursday, October 22

Time – 8:30am – 4:00pm