Nov 17: Matt Taylor at Bryn Mawr

Help an Agent Out: Learning From the Environment and Humans

Matthew E. Taylor, Lafayette College

Thursday, November 17th, 2011 from 4:00-5:00pm

Park 243 (Physics Lecture Hall), Bryn Mawr College

Abstract

---------------------------

Significant advances have been made in autonomous learning, from game playing to training a robot to walk to autonomous helicopter flight. However, we have little understanding about how to best teach such agents. This talk will first present background on reinforcement learning, a paradigm where virtual and robotic agents can autonomously learn to act in complex environments. We will then discuss a selection of recent work towards integrating autonomous learning with advice from other agents or even humans.

Speaker Biography

---------------------------

Matthew E. Taylor graduated magna cum laude with a double major in computer science and physics from Amherst College in 2001. After working for two years as a software developer, he began his Ph.D. with a MCD fellowship from the College of Natural Sciences. He received his doctorate from the Department of Computer Sciences at the University of Texas at Austin in the summer of 2008, supervised by Peter Stone. Matt recently completed a two year postdoctoral research position at the University of Southern California with Milind Tambe and is now an assistant professor at Lafayette College in the computer science department. His current research interests include intelligent agents, multi-agent systems, reinforcement learning, and transfer learning.

Nov 21, 2011: Steve Castellotti at Villanova

The Brain Computer Interface

Steve Castellotti, Puzzlebox, LLC

Monday, November 21, 2011 at 04:30 PM

Mendel Science Center 101

Villanova University

Abstract:

Brain-Computer Interface (BCI) technology has advanced rapidly in recent years, moving out of research labs and into the medical field and consumer space. Applications range from helping the injured and disabled communicate, navigate, and rehabilitate to law enforcement, controlling robots, toys, and playing video games. This presentation will cover basic principles behind the science involved through live demonstrations and examples of the current state of the art. Potential future uses, practical limitations, and various controversies will also be discussed. Following a question and answer session attendees will have an opportunity to try out a variety of BCI mechanisms for themselves first-hand.

Apr 8: Gary McGraw @ Swarthmore

Attack Trends or Why Software Security

Gary McGraw, Cigital
Friday, April 8, 4:30pm
Science Center, Room 199, Swarthmore College

In some sense, software is the lifeblood of most modern complex systems. Software can fail, but worse yet, software can be intentionally made to fail by attackers. Instead of defending our
systems by isolating them from the network (an impossible task), we must build security in from the beginning. Both social networking and mobile device security provide important security lessons that can inform a reasoned approach. Modern malicious code, including the Zeus Trojan, Stuxnet, and other persistent web threats, is as sophisticated as it is insidious. And future trends in attacks are even more alarming, leveraging rootkits, multi-core attacks, and hard-to-diagnose timing issues. Our sole recourse is software security. The good news is that we actually know what to do to build security in.

BIO

Gary McGraw is the CTO of Cigital, Inc., a software security consulting firm with headquarters in the Washington, D.C. area. He is a recognized authority on software security and the author of eight best selling books on this topic. His titles include Java Security, Building Secure Software, Exploiting Software, Software Security, and Exploiting Online Games; and he is editor of the Addison-Wesley Software Security series. Dr. McGraw has also written over 100 peer-reviewed scientific publications, authors a monthly security column for informIT, and is frequently quoted in the press. Besides serving as a strategic counselor for top business and IT executives, Gary is on the Advisory Boards of Fortify Software (acquired by HP), Invincea, and Raven White. His dual PhD is in Cognitive Science and Computer Science from Indiana University where he serves on the Dean's Advisory Council for the School of Informatics. Gary served on the IEEE Computer Society Board of Governors and produces the monthly Silver Bullet Security Podcast for IEEE Security & Privacy magazine (syndicated by informIT).

Mar 23: Jennifer Neville at Bryn Mawr

The Computer Science Department at Bryn Mawr College invites you to a special event: Invited Talk on Machine Learning

Modeling and Mining Social Networks
Jennifer Neville, Assistant Professor, Purdue University
Wednesday, March 23, 2011 from 4:00-5:00 pm
Park 243 (Physics Lecture Hall), Bryn Mawr College

Abstract:
In the past decade, we have witnessed an explosive growth in the use of the Web and online communities. This has lead to increased interest in mining the resulting social network data, both to advance understanding of human behavior and to exploit the underlying social processes for decision-making. In complex network domains (e.g., communication, friendship, and organizational networks), the relationships are a critical source of information that identify potential statistical dependencies among individuals. These dependencies among linked entities present an opportunity to improve predictions about the properties of individuals, as birds of a feather do indeed flock together. For example, when deciding how to market a product to people in Facebook or LinkedIn, it may be helpful to consider whether a person's friends are likely to purchase/adopt the product.

In this talk, to give an overview of the types of statistical learning and inference challenges in network domains, I will present work from three different, yet related, areas of social network mining. First, I will discuss the issue of how to sample a representative subgraph from a large complex network in order to efficiently study domain characteristics and support development of network systems. Next, I will describe a machine learning method to automatically infer relationship strength (e.g., strong vs. weak) from social activity patterns in online social networks, where the goal is to identify influential relationships and prune away spurious links. Finally, I will discuss how to differentiate the behavioral effects of social influence and homophily in networks changing over time and outline a novel statistical test to determine which effects are significant.

Bio:
Jennifer Neville is an assistant professor at Purdue University with a joint appointment in the Departments of Computer Science and Statistics. She received her PhD from the University of Massachusetts Amherst in 2006. She received a DARPA IPTO Young Investigator Award in 2003 and was selected as a member of the DARPA Computer Science Study Group in 2007. In 2008, she was chosen by IEEE as one of "AI's 10 to watch." Her research focuses on developing data mining and machine learning techniques for relational domains, including citation analysis, fraud detection, and social network analysis.

Hal Berghel at Bryn Mawr

Talk: Secure Credentialing
Speaker: Hal Berghel, UNLV
Date: Friday, January 21, 2011, 3:00pm
Location: Park Science Building, Bryn Mawr College

We discuss several new methods for the creation of secure credentials, including some of those for which the speaker holds patents. These methods include those that work with conventional identification media (mag stripe cards, smart cards, RFID cards, etc.) as well as newer applications that use digital displays (e.g., on iPhones and PDAs). These methods will be presented in the context of a variety of business, government, law enforcement and military applications. Our methods integrate biometrics (fingerprint, iris scan, bone scan, capillary/palm scan, photographic images, etc.) to provide at least four points of authentication. Industry standard encryption (e.g., AES and Blowfish) is added in a variety of ways to provide security. The result is a self-validating credential that operates on a mobile platform with equipment that may be found in most office equipment retail stores. One of our systems, CardSleuth, will be demonstrated. Although CardSleuth takes advantage with elecrical power and network access, it requires neither for full functionality. The software runs on any Windows computer, PDA, phone, etc. for both the generation and recognition, as well as authentication and validation of IDs. The robustness of these methods are compared with recent government efforts such as RealID and the WHTI Pass Card.

Hal Berghel is currently Professor and Director of the School of Informatics at the University of Nevada, Las Vegas where he has previously served as Director of the School of Computer Science and Associate Dean of the College of Engineering. He is also the founding Director of both the Center for CyberSecurity Research, and the Identity Theft and Financial Fraud Research and Operations Center. His research interests are both catholic and eclectic, ranging from logic programming and expert systems, relational database design, algorithms for non-resolution based inferencing, approximate string matching, digital watermarking and steganography, and digital security (including both computer and network forensics), For the past decade he has applied his work in digital security to law enforcement, particularly with respect to digital crime, cyberterrorism, and information warfare. His research has been supported by both industry and government for over thirty years. His current in secure credentialling technology is funded by the Department of Justice. In addition to his academic positions, Berghel is also a popular columnist, author, frequent, talk show guest, inventor, and keynote speaker. For nearly fifteen years he wrote the popular Digital Village column for the Communications of the ACM.

Berghel is a Fellow of both the Institute for Electrical and Electronics Engineers and the Association for Computing Machinery, and serves both societies as a Distinguished Lecturer. He has received the ACM Outstanding Lecturer of the Year Award four times and was recognized for Lifetime Achievement in 2004. He has also received both the ACM Outstanding Contribution and Distinguished Service awards. He is also the founder and owner of Berghel.Net, a consultancy serving business and industry, and co-owner of BC Innovations Management, a startup company in IP and DRM.