Wednesday, May 17, 7:30-9:30AM

Title:

A Practical Introduction to Applied Machine Learning For Scientists

Abstract:

We've seen the rapid adoption of machine learning in many application areas recently. These include computer vision, speech recognition, biometrics, natural language processing, fin-tech, e-commerce, agri-tech, manufacturing, automotive, health-tech, biopharma, and, yes, in aerospace and space exploration. Machine learning is the underpinning of numerous advances in so called “artificial intelligence” and may rival electricity or the transistor as one of the most transformative technology catalysts the world has experienced in the last 200 years. All the hype notwithstanding, a good question technicians and scientists should ask themselves - Is machine learning a useful tool to further advance their work or research?

In this workshop, we will cut through the hype and give participants hands-on experience in several key machine learning areas including:

• Preparing data for machine learning using exploratory data analysis
• Training a model for time-series based forecasting
• Evaluating the model for accuracy
• Optimizing the model for speed and size
• Deploying the model to custom accelerator hardware
• Monitoring the model in production

We will use open-source tools such as Python and PyTorch and state-of-the-art machine learning techniques including deep learning and sequence modeling. Participants need only bring a laptop with an internet browser. The workshop will provide all the tools and the dataset. By the end of the 2-hour workshop, participants will have gained practical experience with the 360 degree lifecycle of machine learning- from data preparation, to model training, optimization, deployment, and performance monitoring. Workshop materials will be made available before the workshop for participants who want to get a head-start.

Biography:

George Williams is the Head of AI at Smile Identity, an identity management and computer vision-based biometrics provider. He has held senior leadership roles in software engineering, system design, data science, and AI research, including tenures at Apple's New Product Architecture group and at New York University's Courant Institute. He can talk on a broad range of topics at the intersection of e-commerce, machine learning, cybersecurity, computer hardware, and computer science. He is an author of several research papers in computer vision and deep learning, published at NeurIPS, CVPR, ICASSP, ICCV, and SIGGRAPH. George is regularly invited to present at meetups and technology conferences, including recent talks at Blackhat, Open Data Science Conference, Apache Spark Summit, JupyterCon, AnacondaCon, and Space Computing. He is a track chair at the Valleyml.ai conference and as a chair for the Neural Information Processing Conference.

Wednesday, May 17, 10:30-11:00AM

Title:

System-Level Probabilistic Risk Assessment for Single Event Effects

Abstract:

This tutorial presents a Probabilistic Risk Assessment (PRA) methodology to predict the unavailability of parts and systems in a radiation environment using probabilistic models for radiation-induced failure modes, including time-dependent single-event effects (SEE), and their downstream effects. Fault Tree Analysis (FTA) propagates event-level risk using Bayesian networks and custom Monte Carlo solution algorithms. The analysis enables model-based Radiation Hardness Assurance (RHA) that is more flexible and better at quantifying risk than the traditional approach. The PRA technique is demonstrated with a system of real parts, including a Samsung Flash memory, an Elpida SDRAM, and a Xilinx FPGA. Published radiation test data is digested into SEE rates for the target environment using industry-standard tools and models. Expected SEE rates are then combined with system design choices such as watchdog timers, memory utilization, scrub rates, and fault tolerance to predict the unavailability due to each failure mode using custom models. Different mitigation strategies are examined in terms of their improvements to system-level availability.

Biography:

Stephen Lawrence is a recent graduate of the University of Tennessee at Chattanooga, where he completed his M.S. degree in Electrical Engineering with a focus on radiation effects. His research of radiation effects in COTS systems has been molded over 3+ years of hands-on testing experience through university projects and commercial space internships.

The 2023 SEE/MAPLD Combined Workshop is pleased to host the following invited speakers:

Tuesday, May 16th, 2023
2:30 PM	30 Min	Spot the Differences: Planning and Execution of Pulsed-Laser and Heavy-Ion SEE Experiments	Joel Hales	US Naval Research Laboratory
Wednesday, May 17th, 2023
9:30 AM	30 Min	Reaching the Good, and Avoiding the Bad, When Using AI	David Danks	University of CA, San Diego
2:00 PM	20 Min	Microchip RISC-V/FPGA	Tim Morin	Microchip
Thursday, May 18th, 2023
2:30 PM	30 Min	Is an Onboard Artificial-Intelligence (AI) Approach Suitable for Your Science Application?	Justin Goodwill	NASA Goddard Space Flight Center