NEREN Seminar –Friday, October 4, 2019, 9:00AM-2:15PM, Gateway City Arts Center, Holyioke, MA

In collaboration with UMass Amherst and the Massachusetts Green High-Performance Computing Center (MGHPCC), NEREN presents the seventh in a series of day-long seminars devoted to proposing and advancing ideas for regional collaboration in research computing and networking. The theme of the seminar will be AI and Machine Learning. By attending in person, you will gain the added benefit of interpersonal networking with peers from across the Northeast region.

Please register by September 23, 2019, to attend in person or remotely by contacting Laurie Robinson, NEREN Program Administrator, at laurie@neren.org or by phone: 401-523-5107. When registering, please indicate whether you are attending in person or remotely.

AGENDA – Friday, October 4, 2019

9:00 – 9:45 a.m. Registration/Continental Breakfast/Networking

9:45 – 10:00 a.m. Welcoming/Opening Remarks by NEREN, Inc.

Please note that the Webcast/Phone options will be available for the presentations. If joining by webcast, please use the following link:

http://demo.mediasite.oshean.org/Mediasite/Play/02d6533f6dd442f796f4624c250c58461d

10:00 – 10:40 a.m. Presentation #1 –“How to manage the complex implications of face recognition technology: A possible way forward”

Presenter: Erik G. Learned-Miller, Professor in the College of Information and Computer Sciences, University of Massachusetts Amherst

There has been a great deal of hand-wringing about face recognition technology. Important work has shown that face recognition algorithms can be unfair, can amplify pre-existing biases, and can create substantial harm to individuals. Others have testified to the improvements in efficiency of doing important work like tracking down child sex traffickers. Still others argue that many applications of face recognition technology are benign and that large scale bans are unreasonable. How can we integrate and balance these concerns? Many arguments center around machine learning ideas like unbiased learning, better training sets, and algorithms that are robust to “domain transfer”.

In this talk, I will argue that the problem is much larger than these (important) technical issues. To do so, I will examine some of the regulatory structures, processes, definitions, rules, and conventions that have been developed by the US Food and Drug Administration (FDA). I will draw heavily from two separate scenarios: the FDA’s regulation of pharmaceuticals and their regulation of medical devices. The elaborate processes set up to regulate the drug and medical device industries have been remarkably successful (in many ways), and I will argue that many of the structures in place there deserve analogous systems for the regulation of face recognition.

10:40 – 11:20 a.m. Presentation #2 –“Leveraging High Performance GPU Computing to Increase Research Productivity in Vermont”

Presenter: Adrian Del Maestro, Assistant Professor of Physics, University of Vermont

The increasing complexity of scientific data-driven workflows has led to an evolution in the types of tools both employed and demanded by interdisciplinary researchers. In this talk I will describe the deployment of a new GPU based supercomputer at the University of Vermont and detail how we are engaging with a new class of users that are interested in employing machine learning in their research, but may have minimal previous experience with high performance computing. The results highlight that specialized cyberinfrastructure with advanced artificial intelligence capabilities will play an increasing and fundamental role in the university research ecosystem and should be supported at the same level as conventional brick-and-mortar scientific facilities.

11:20 – 12:00 p.m. Presentation #3 – “Hypersparse Neural Network Analysis of Large-Scale Internet Traffic”

Presenter: Jeremy Kepner, MIT Lincoln Laboratory Fellow

The Internet is transforming our society, necessitating a quantitative understanding of Internet traffic. Our team collects and curates the largest publicly available Internet traffic data containing 50 billion packets. Utilizing a novel hypersparse neural network analysis of "video" streams of this traffic using 10,000 processors in the MIT SuperCloud reveals a new phenomena: the importance of otherwise unseen leaf nodes and isolated links in Internet traffic. Our neural network approach further shows that a two-parameter modified Zipf-Mandelbrot distribution accurately describes a wide variety of source/destination statistics on moving sample windows ranging from 100,000 to 100,000,000 packets over collections that span years and continents. The inferred model parameters distinguish different network streams and the model leaf parameter strongly correlates with the fraction of the traffic in different underlying network topologies. The hypersparse neural network pipeline is highly adaptable and different network statistics and training models can be incorporated with simple changes to the image filter functions.

12 noon – 1:10 p.m. Lunch Break

1:15 – 2:00 p.m. Presentation #4 – “Knowledge Representation in the Era of Deep Learning, Watson and the Semantic Web”

Presenter: Jim Hendler, Tetherless World Chair of Computer, Web and Cognitive Sciences, Director RPI/IBM AI Research Collaboration, Rensselaer Polytechnic Institute (RPI)

A burst in optimism (and unwarranted fear) has grown around a number of technologies that are high impact and able to solve problems that have challenged AI researchers for years. The over-enthusiasm that often follows such breakthroughs has caused some to declare (yet again) that it is the end of “knowledge representation” as AI moves into a world dominated by neural networks, data mining and the knowledge graph. In this talk, I argue that these technologies, while extremely powerful separately, are not only still a long way from human intelligence, but cannot get there without a level of knowledge and reasoning beyond what is currently available to these techniques, On the other hand, I also argue that taking these technologies into new and harder realms will require rethinking what traditional AI representation is and how it is used.

2:00 – 2:10 p.m. Closing Remarks: Christopher Misra, Vice Chancellor and CIO, University of Massachusetts Amherst

To register, or if you have questions, please email Laurie Robinson, NEREN Program Administrator, at laurie@neren.org or by phone: 401-523-5107. Or register online:

GETTING THERE – DETAILS

Getting There in Person

The following link provides details about the location for those traveling to the meeting by car:

https://www.google.com/maps/dir/41.608711,71.455757/gateway+city+arts+center+directions+holyoke/@41.9088149,72.5213322,9z/data=!3m1!4b1!4m9!4m8!1m1!4e1!1m5!1m1!1s0x89e6dc180714a031:0x78d 687c11d3e341b!2m2!1d-72.6038625!2d42.2042256

Helpful Telephone Numbers/Contact Information

John Griffin, (UMass Amherst) 413-545-9939 or jgriffin@umass.edu

John Goodhue (MGHPCC) 617-834-5601 or jtgoodhue@mghpcc.org

Laurie Robinson, (NEREN) 401-523-5107, laurie@neren.org

For Technical Questions: Jim Carr, Consultant, (OSHEAN), 401-447-5600, jim@oshean.org

Meeting Location/Wireless Internet Access

Gateway City Arts Center does have a guest wireless network.

Parking/Public Transportation

Street parking is available, and parking is also available at the MGPHCC, 100 Bigelow Street, Holyoke, which is located within short walking distance to Gateway City Arts Center, 92 Race Street, Holyoke, Massachusetts.

Getting There Remotely – Phone/Webcast (Watch for details on how to Join)

Please note that the Webcast will be recorded, and we will provide details following the seminar. Thank you to David Marble, President and CEO of OSHEAN for assisting with the Webcast/Video Steaming!