Deep Conversations on Deep Learning: A technical discussion series III
Deep Conversations on Deep Learning: A technical discussion series
Interesting conversations of the latest topics in Artificial Intelligence and Machine Learning
Organized by the IEEE Maine Section with presentation resources provided by IEEE Region 1
Organized and hosted by Walter Rawle, Chair, IEEE Maine Section
Occurs the third Wednesday of every month effective 9/16/2020 until 12/16/2020 from 6:00 PM to 7:00 PM, (UTC-04:00) Eastern Time (US & Canada)
Date and Time
Location
Hosts
Registration
- Date: 18 Nov 2020
- Time: 06:00 PM to 07:00 PM
- All times are (GMT-05:00) US/Eastern
- Add Event to Calendar
- Via WebEx
- Augusta/Bangor/Portland, Maine
- United States
- Starts 10 September 2020 09:00 AM
- Ends 18 November 2020 07:00 PM
- All times are (GMT-05:00) US/Eastern
- No Admission Charge
Agenda
Deep Conversations on Deep Learning, a technical discussion series, has been organized by the IEEE Maine Section to provide IEEE members technical insight into some of the fascinating emerging developments in the field. Mixing both fundamental concepts and application discussions, the series is intended to spur interest and motivate IEEE members to explore this area further. Presentations have been scheduled for the third Wednesday of each month, September through December, at 6 pm local eastern time, so that members who hold day jobs can participate in the events after work hours. Slide decks will be available on the IEEE Maine Section website after each presentation and the presenters will be available for fifteen minutes after each presentation to answer questions. The IEEE Maine section hopes that this series will be of value to its members and looks forward to suggestions on future technical series presentations. Presentation schedules and topic abstracts are as follows:
The Deep Learning Data Dilemma: Solving It with Physics Based Modeling: November 18, 6 pm: Presenter: Walter Rawle
Abstract: It is well known that machine learning requires lots of data. Lots of data!! Data is required to train networks. Data is required to test networks. Data is required to optimize networks. And “ground truth” data, actual measurements, is expensive to obtain. A potential solution to the expense of gathering data for machine learning training is simulation developed with physics-based models (PBM). This talk will focus on the development of physics-based model simulations and discuss two examples, from recent work, where PBMs have been employed for 1) mechanical diagnostics and 2) LIDAR perception systems. The mechanical diagnostics example discusses the acquisition of simulated “ground truth” for helicopter drivetrains. The LIDAR example discusses elastic backscatter correction, incorporating both molecular absorption and stochastic physical scattering. This talk has been designed for a more analytical, technical audience. However, IEEE Maine hopes that its members find this topic informative and interesting.