IBM and NASA worked with SETI and NVIDIA to run several AI driven planetary defense and space weather projects in the summer of 2017. Learn about the technologies used and how these groups are working to better understand the protection of our planet.

1. Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
Predicting Space Weather and
Planetary Defense with NASA FDL and
IBM Cloud
—
Naeem Altaf
Distinguished Engineer
Brad DesAulniers
Sr. Solution Architect

2. Please note
IBM’s statements regarding its plans, directions, and intent are subject to change
or withdrawal without notice and at IBM’s sole discretion.
Information regarding potential future products is intended to outline our general
product direction and it should not be relied on in making a purchasing decision.
The information mentioned regarding potential future products is not a commitment, promise,
or legal obligation to deliver any material, code or functionality. Information about potential
future products may not be incorporated into any contract.
The development, release, and timing of any future features or functionality described for our
products remains at our sole discretion.
Performance is based on measurements and projections using standard IBM benchmarks in
a controlled environment. The actual throughput or performance that any user will
experience will vary depending upon many factors, including considerations such as the
amount of multiprogramming in the user’s job stream, the I/O configuration, the storage
configuration, and the workload processed. Therefore, no assurance can be given that an
individual user will achieve results similar to those stated here.
2

3. Notices and disclaimers
3Think 2018 / January 12, 2018 / © 2018 IBM Corporation
© 2018 International Business Machines Corporation. No part of this
document may be reproduced or transmitted in any form without
written permission from IBM.
U.S. Government Users Restricted Rights — use, duplication or
disclosure restricted by GSA ADP Schedule Contract with IBM.
Information in these presentations (including information relating to
products that have not yet been announced by IBM) has been reviewed
for accuracy as of the date of initial publication and could include
unintentional technical or typographical errors. IBM shall have no
responsibility to update this information. This document is distributed
“as is” without any warranty, either express or implied. In no event,
shall IBM be liable for any damage arising from the use of this
information, including but not limited to, loss of data, business
interruption, loss of profit or loss of opportunity. IBM products and
services are warranted per the terms and conditions of the agreements
under which they are provided.
IBM products are manufactured from new parts or new and used parts.
In some cases, a product may not be new and may have been previously
installed. Regardless, our warranty terms apply.”
Any statements regarding IBM's future direction, intent or product
plans are subject to change or withdrawal without notice.
Performance data contained herein was generally obtained in a
controlled, isolated environments. Customer examples are presented as
illustrations of how those
customers have used IBM products and the results they may have
achieved. Actual performance, cost, savings or other results in other
operating environments may vary.
References in this document to IBM products, programs, or services does
not imply that IBM intends to make such products, programs or services
available in all countries in which IBM operates or does business.
Workshops, sessions and associated materials may have been prepared
by independent session speakers, and do not necessarily reflect the
views of IBM. All materials and discussions are provided for informational
purposes only, and are neither intended to, nor shall constitute legal or
other guidance or advice to any individual participant or their specific
situation.
It is the customer’s responsibility to insure its own compliance with legal
requirements and to obtain advice of competent legal counsel as to
the identification and interpretation of any relevant laws and regulatory
requirements that may affect the customer’s business and any actions
the customer may need to take to comply with such laws. IBM does not
provide legal advice or represent or warrant that its services or products
will ensure that the customer follows any law.

4. Notices and disclaimers
continued
4Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
Information concerning non-IBM products was obtained from the
suppliers of those products, their published announcements or other
publicly available sources. IBM has not tested those products about this
publication and cannot confirm the accuracy of performance, compatibility
or any other claims related to non-IBM products. Questions on the
capabilities of non-IBM products should be addressed to the suppliers of
those products. IBM does not warrant the quality of any third-party
products, or the ability of any such third-party products to
interoperate with IBM’s products. IBM expressly disclaims all
warranties, expressed or implied, including but not limited to, the
implied warranties of merchantability and fitness for a purpose.
The provision of the information contained herein is not intended to, and
does not, grant any right or license under any IBM patents, copyrights,
trademarks or other intellectual property right.
IBM, the IBM logo, ibm.com and [names of other referenced IBM
products and services used in the presentation] are trademarks of
International Business Machines Corporation, registered in many
jurisdictions worldwide. Other product and service names might
be trademarks of IBM or other companies. A current list of IBM
trademarks is available on the Web at "Copyright and trademark
information" at: www.ibm.com/legal/copytrade.shtml.
.

5. NASA Frontier
Development Lab and
IBM
Improving planet
safety with AI
5Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation

6. 6Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
• NASA FDL runs an intense 8-week concentrated study program
hosted at the SETI Institute and NASA Ames to address space
problems using advanced DL / AI techniques
• Over 40 post-doc researchers and professional scientists
participated to NASA Ames tackle these problems using the IBM
provisioned environment:
• Lockheed Martin Solar Physics Division
• CERN AI Research, NASA JPL, NASA Ames
• Oxford University, Stanford, Cornell, Caltech, Cambridge,
Georgia Tech, etc.
NASA FRONTIER DEVELOPMENT LAB
An applied artificial intelligence research accelerator established to maximize
new AI technologies and capacities emerging in academia and the private
sector and apply them to challenges in space

7. PLANETARY DEFENSE
LONG PERIOD COMETS
PROVIDE MORE WARNING TIME FOR LONG-
PERIOD COMET IMPACTS BY APPLYING DEEP
LEARNING TO METEOR SHOWER
OBSERVATIONS.
7Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
RADAR 3D SHAPE MODELING
DEVELOP A METHODOLOGY TO AUTOMATE
THE BACKLOG OF NEO RADAR IMAGERY
THAT REQUIRES SHAPE MODELING - AND
ALSO IMPROVE THE RESOLUTION OF THE
RESULT.

8. SPACE WEATHER
SOLAR-TERRESTRIAL INTERACTIONS
IMPROVE UNDERSTANDING OF SOLAR
INFLUENCE ON EARTH’S MAGNETOSPHERE
AND ATMOSPHERE.
8Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
SOLAR STORM PREDICTION
DISCOVERING NEW RELATIONSHIPS
AND AGENTS TO HELP PREDICT MAJOR
SOLAR EVENTS.

9. LONG-PERIOD COMETS
PROVIDE MORE WARNING
TIME FOR LONG-PERIOD
COMET IMPACTS BY
APPLYING DEEP LEARNING
TO METEOR SHOWER
OBSERVATIONS.
9Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
Add years of extra warning time to finding the
otherwise hard-to-spot dangerous space objects.
This task is particularly suited to a machine learning
approach because of the large scale of data, the
need for integration of surveys from around the
globe without human intervention, and the need to
operate for a long period of time.
The deep learning algorithms would be used to
recognize meteors amongst false positives (e.g.,
satellites), and can triangulate the meteor trajectory
in Earth’s atmosphere, its entry speed, and the pre-
impact orbit in space through combining different
camera objectives to the same meteor.
Planetary Defense

10. LONG-PERIOD COMETS
Technologies Used
10Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
• Long Short Term Memory (LSTM)
• Special flavor of Recurrent Neural Networks
(RNNs)
• Precision 90%
• Recall 89%
• Convolutional Neural Network (CNN)
• Standard AlexNet architecture
• Precision 88.6%
• Recall 90.3%
• Radio Frequency Machine Learning Systems
(RFMLS)
• Trained a random forest to classify meteor vs
non-meteor in dataset of ~200,000 objects
from CAMS
• Precision: 90%
• Recall: 80.6%
Planetary Defense

11. Results – Planetary Defense (Long-Period
Comets)
11Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
• Improved and automated the identification of meteors above human level
performance on images detected in meteor shower surveys using machine
learning and deep learning
• Recovered known meteor shower streams and characterized previously unknown
meteor showers from meteor orbital data
• We are able to find rare meteor outbursts from dust trail encounters that could
trace the orbits of dangerous long-period comets

12. RADAR 3D SHAPE
MODELING
INVERTING RADAR
IMAGES: SCALING
RESOLUTION AND
AUTOMATION OF SHAPE
MODELING NEAR EARTH
OBJECTS.
12Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
During FDL 2016 the radar shape modeling team
studied ways to accelerate and better automate
analysis of delay-Doppler radar images of asteroids
obtained by the Arecibo and Goldstone planetary
radars. Bayesian optimization and VAE (Variational
Auto Encoder) neural net approaches were
implemented showing promising results. The VAE
approach in particular suggests further opportunity
for producing more detailed shape models from
delay-Doppler data and identifying specific features.
First Objective: To investigate how the performance
of the VAE network architecture scales as the
number of latent variables is increased.
Second: To investigate other architectures used in
machine vision and remote 3D modeling
applications and determine if they have additional
value to planetary defense.
Planetary Defense

13. RADAR 3D SHAPE
MODELING
Technologies Used
13Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
• Starts with Raw Radar Data
• Preprocessing generates a simplified shape
• Bayesian optimization to automate pole
searches (spin axis direction)
• Deep Neural Network (DNN)
• Generative adversarial networks (GANs)
• 3D Representations generated from output of
DNNs and GANs using Point Clouds in Graphs
Planetary Defense

14. Results – Planetary Defense (3D Shape
Modeling)
14Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
• Pre-processing is faster
• Spin axis determination is faster
• Training data generation is improved
• Neural network is improved

15. SOLAR-TERRESTRIAL
INTERACTIONS
IMPROVE
UNDERSTANDING OF
SOLAR INFLUENCE ON
EARTH’S
MAGNETOSPHERE AND
ATMOSPHERE.
15Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
The term ‘Space Weather’ generally refers to
conditions on the sun and in the solar wind,
interacting with our planet’s magnetosphere and
upper atmosphere and in turn impacting the
performance and reliability of space and ground
based technological systems.
The opportunity offered by AI techniques is remote
sensing and in situ measurements from NASA’s
Heliophysics System Observatory exploring the
connections between solar forcing, heliospheric
changes and manifestations of space weather in the
Earth’s magnetosphere and atmosphere.
Potential breakthroughs include finding evidence for
solar influence on lightning patterns, major
improvements in the ability to predict the duration
and strength of geomagnetic storms, and the ability
to predict ionospheric changes during major solar
storms.
Space Weather

16. SOLAR-TERRESTRIAL
INTERACTIONS
Technologies Used
16Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
• B-Sting: Solar Terrestrial Interactions Neural
Network Generator
• OpenSource data-driven tool for space
weather forecasting
• Predicts Kp (common index that captures
geomagnetic disturbances)
• Keras on top of Tensorflow
• Long Short Term Memory (LSTM)
• Special flavor of Recurrent Neural Networks
(RNNs) capable of learning long term
dependencies
• Able to predict Kp 3-6 hours ahead with > 95%
confidence
Space Weather

17. Results – Solar-Terrestrial Interactions
17Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
• Prediction of Kp 3 hours ahead
• Confidence level of more than 95%

18. SOLAR STORM
PREDICTION
DISCOVERING NEW
RELATIONSHIPS AND
AGENTS TO HELP PREDICT
MAJOR SOLAR EVENTS.
18Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
There have been 26 significant solar storm events
affecting Earth over the last 50 years. In serious
cases, these solar events can cause significant
damage to human infrastructure. Solar storms are
particularly significant as we begin to think about
moving out of LEO into deep space exploration,
permanently crewed facilities on the Moon, cis-lunar
space and NASA’s goal to visit Mars within the next
two decades.
Emerging AI tools offer the opportunity to analyze
variations in the solar magnetic field and solar
corona using data from the Solar Dynamics
Observatory (SDO, surface vector magnetograms,
and EUV images), in order to discover relationships
between the observed magnetic activity in the
photosphere and corona and to identify the agents
that drive solar eruptive events (flares and coronal
mass ejections).
Space Weather

19. SOLAR STORM
PREDICTION
Technologies Used
19Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
• Convolutional Neural Networks (CNNs)
• Data from Solar Dynamics Observatory (SDO)
• One of the SDO instruments, the Atmospheric
Imaging Assembly (AIA), focuses on the
evolution of the magnetic environment in the
Sun’s atmosphere and its interaction with
embedded and surrounding plasma
• FlareNet
• https://github.com/nasa-fdl/flarenet
• Components for downloading and
transforming SDO data, specifying network
architectures, and running experiments
Space Weather

20. Results – Solar Storm Team
20Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
• Dimensionality Reduction
• Topological model of solar magnetic field
• Trained Convolutional Neural Network
• A flare forecasting tool (FlareNet)

21. IBM Cloud Cognitive Systems & Solution Architecture
21Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
Object
Storage
Staging DataIngesting Data
Aspera
Preparing Data
Processing
Model Training / Testing /
Validation/ Inferencing
21
5
6
3
4
16 bare-metal servers, each with:
•CPU = Dual Xeon E5-2690v4 / 28
cores / 2.60 GHz
•GPU = 2 x Tesla P100 with 16GB
(PCIe form factor)
•Memory = 128GB
•Network: 10GBps dual NIC / 10
Gbps
•Local Disk: 1.2 TB SSD (10 DWPD)
•OS: Ubuntu Linux 16.04 LTS Xenial
Xerus
• ACE
• HMI
• AIA
• Lightning Data
• Radar Images
1. RAW data from source moved via High Speed Ingest “Aspera”
2. RAW data Stored in “Cloud Object Storage”
3. RAW data Accessed by Notebooks using spark and python libraries, Data is Processed using “Data Science Experience”
4. Processed data stored back in “Cloud Object Storage”
5. Processed data fetched in Chunks by “Cognitive Systems – GPU Enabled” to local disks, for Training/Testing/Validating
the Model and Inferencing
6. Final results (Predictions) are written to local disk and copied over back to “Cloud Object Storage”

22. IBM NASA FDL 2017 Results
22Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
Benefits and Key Metrics
• A team of researchers and data scientists used machine learning techniques on the IBM Cloud to:
• Develop new processes for 3D modeling of asteroids from radar data
• Developed a solar flare forecasting model capable of warning stakeholders at least an hour in advance
Solution & Technology
• IBM Cloud Cognitive Systems with NVIDIA P100 GPUs
• IBM Cloud Object Storage
• IBM Data Science Experience
Press & Publication
• CBR: https://www.cbronline.com/news/ibm-iot-rotterdam-nvidia-gpu-cloud
• Fast Company: https://www.fastcompany.com/40498881/nasa-silicon-valley-ai-frontier-development-lab
• HPCwire: https://www.hpcwire.com/2018/01/31/ibm-adds-v100-gpus-ibm-cloud-targets-ai-hpc/
• IBM Blog post: https://www.ibm.com/blogs/cloud-computing/2018/01/ai-hpc-workloads-nvidia/
• TechRepublic: https://www.techrepublic.com/article/nvidia-brings-its-fastest-gpu-accelerator-to-ibm-cloud-to-boost-ai-hpc-
workloads/
• FlareNet: A Deep Learning Framework for Solar Phenomena Prediction:
https://dl4physicalsciences.github.io/files/nips_dlps_2017_5.pdf

23. IBM NASA FDL 2018: What’s Next !
23Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
NASA FDL Program: June 25th – August 17th 2018
Growing depth and breadth and impact
Science domains has grown from 3 to 6
FDL Research teams have grown from 5 to 8
Science challenges from NASA have grown from 9 to 12
The Government’s Space Policy Directive (signed Dec 2017) puts this work
at the center of NASA’s immediate science priorities and mission goals
Projects
Space Weather
Earth Observation Team
Technology
Data Science Experience
Watson Machine Learning
PowerAI Cognitive Systems
IBM Q (Quantum Compute)
NASA AI
NASA is actively establishing its AI enhancement plans for the NASA
Advanced Supercomputer (Pleiades) installation in Mountain View, and
participate in the FDL program to determine infrastructure requirements and
vendor strengths

24. References
24Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
IBM and FDL
NASA FDL
SETI
SETI 2
FlareNet for Solar Prediction
HPC Workload Blog

25. Thank you
25Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation
Naeem Altaf
Distinguished Engineer
—
naltaf@us.ibm.com
+1-555-555-5555
ibm.com
Brad DesAulniers
Solution Architect
—
bradd@us.ibm.com
+1-555-555-5555
ibm.com

26. 26Think 2018 / DOC ID / Month XX, 2018 / © 2018 IBM Corporation