ML@GT Distinguished Seminar by Pedro Domingo (UW) on “Sum-Product Networks: The Next Generation of Deep Models” 4/19/2017 @ 12n EBB
Title: “Sum-Product Networks: The Next Generation of Deep Models”
Speaker: Pedro Domingo (University of Washington)
Date/Time: April 19, 2017, @ 12n (lunch served at 11:30am)
Location: Engineered Biosystems Building (EBB), CHOA Room
Abstract: The two main types of deep learning are function approximation and probability estimation. Function approximators like convolutional neural networks are robust and allow for real-time inference, but are very inflexible, requiring fixed inputs and outputs and detailed supervision. Probability estimators like deep Boltzmann machines allow arbitrary inputs and outputs and require no supervision, but are not robust and do not allow real-time inference.
Both are very opaque. Sum-product networks (SPNs) are a new class of deep models that are suitable for both function approximation and probability estimation. SPNs allow for real-time inference, are robust and comprehensible, and are highly flexible, with any choice of inputs and outputs and any amount of supervision. I will present generative and discriminative algorithms for learning SPN weights, and an algorithm for learning SPN structure. SPNs have achieved impressive results in a wide variety of domains, including object recognition, image completion, activity recognition, language modeling, collaborative filtering, and click prediction, and are arguably the most powerful class of deep models available today. (Joint work with Abe Friesen, Rob Gens, Mathias Niepert and Hoifung Poon.)
Bio: Pedro Domingos is a professor of computer science at the University of Washington and the author of “The Master Algorithm”. He is a winner of the SIGKDD Innovation Award, the highest honor in data science. He is a Fellow of the Association for the Advancement of Artificial Intelligence and has received a Fulbright Scholarship, a Sloan Fellowship, the National Science Foundation’s CAREER Award, and numerous best paper awards. His research spans a wide variety of topics in machine learning, artificial intelligence, and data science, including scaling learning algorithms to big data, maximizing word of mouth in social networks, unifying logic and probability, and deep learning.
Devi Parikh (IC) Awarded the Prestigious 2017 IJCAI Computers and Thought Award
Congrats to Devi Parkih of the School of Interactive Computing (IC) for being awarded the prestigious IJCAI Computers and Thought Award for 2017. This award is presented every two years by International Joint Conferences on Artificial Intelligence (IJCAI) and is considered to be “the premier award for artificial intelligence researchers under the age of 35” to recognize outstanding young scientists in artificial intelligence. It was originally funded with royalties received from the book Computers and Thought (edited by Edward Feigenbaum and Julian Feldman), and is currently funded by IJCAI.
- See this Wikipedia entry for names of past winners. An impressive list.
Welcome The Agency, the undergraduate AI/ML club at Georgia Tech
The Agency is an undergraduate AI/ML club at Georgia Tech. Here is a short introduction to The Agency, from the members of this club.
We are a hub for connecting undergraduate students with the machine learning community at Tech. Every week, we present talks on machine learning/artificial intelligence. These range from undergraduates lecturing on interesting concepts in the field to professors/graduate students presenting their research.
Beyond our weekly lectures, we give people the resources to do machine learning research. For instance, we organize larger events, such as our recent workshop on TensorFlow, and offer dedicated computing resources for undergraduates to train learning models. We also host a deep learning paper reading group that meets every two weeks.
Furthermore, we also work on projects such as Buzzmobile, an autonomous vehicle modeled after the Rambling Wreck. Buzzmobile has a purely reactive control architecture, built entirely in ROSpy. Other shorter-term projects we’ve worked on in the past include various game AIs and applications of different AI and ML algorithms and models, such as RNNs for lyrics generation or simulated annealing for prettifying graphs.
Many of our members and officers are already doing undergraduate research in the field of Machine Learning. We are hoping to grow this community of undergraduate researchers by providing our members with the resources of and connections to ML labs (or labs that can use ML).
We welcome The Agency to ML@GT family.
ML@GT Seminar by Le Song (CSE) on “Embedding as a Tool for Algorithm Design” on April 5, 2017, 12:00n in EBB CHOA Room
Speaker: Le Song, Computational Science and Engineering (CSE), GA Tech
Date/Time: April 5, 2017, 12:00n – 1:00pm (Lunch at 11:30am)
Title: Embedding as a Tool for Algorithm Design
Abstract: Many big data analytics problems are intrinsically complex and hard, making the design of effective and scalable algorithms very challenging. Domain experts need to perform extensive research, and experiment with many trial-and-errors, in order to craft approximation or heuristic schemes that meet the dual goals of effectiveness and scalability. Very often, restricted assumptions about the data, which are likely to be violated in real world, are made in order for the algorithms to work and obtain performance guarantees. Furthermore, previous algorithm design paradigms seldom systematically exploit a common trait of real-world problems: instances of the same type of problem are solved repeatedly on a regular basis, differing only in their data. Is there a better way to design effective and scalable algorithms for big data analytics?
I will present a framework for addressing this challenge based on the idea of embedding algorithm steps into nonlinear spaces, and learn these embedded algorithms from problem instances via either direct supervision or reinforcement learning. In contrast to traditional algorithm design where every step in an algorithm is prescribed by experts, the embedding design will delegate some difficult algorithm choices to nonlinear learning models so as to avoid either large memory requirement, restricted assumptions on the data, or limited design space exploration. I will illustrate the benefit of this new design framework using large scale real world data, including a materials discovery problem, a recommendation problem over dynamic information networks, and a problem of learning combinatorial algorithms over graphs. The learned algorithms can reduce memory usage and runtime by orders of magnitude, and sometimes result in drastic improvement in predictive performance.
Bio: Le Song is an Associate Professor in the Department of Computational Science and Engineering, College of Computing, and an Associate Director of the Center for Machine Learning, Georgia Institute of Technology. He received his Ph.D. in Machine Learning from University of Sydney and NICTA in 2008, and then conducted his post-doctoral research in the Department of Machine Learning, Carnegie Mellon University, between 2008 and 2011. Before he joined Georgia Institute of Technology in 2011, he was a research scientist at Google briefly. His principal research direction is machine learning, especially nonlinear models, such as kernel methods and deep learning, and probabilistic graphical models for large scale and complex problems, arising from artificial intelligence, network analysis, computational biology and other interdisciplinary domains. He is the recipient of the Recsys’16 Deep Learning Workshop Best Paper Award, AISTATS’16 Best Student Paper Award, IPDPS’15 Best Paper Award, NSF CAREER Award’14, NIPS’13 Outstanding Paper Award, and ICML’10 Best Paper Award. He has also served as the area chair for many leading machine learning and AI conferences such as ICML, NIPS, AISTATS, AAAI and IJCAI, and the action editor for JMLR.
ML@GT Launch and Welcome Celebration, April 17, 2017 11am-5pm, TSRB Auditorium
The Machine Learning Center at Georgia Tech (ML@GT) invites you to the ML@GT Welcoming Celebration for faculty and students from across campus who are engaged in the field of Machine Learning. Come listen to those in the ML community speak about new research and how ML is changing our world. The day will feature talks from our faculty and a poster session by our students showcasing our best work.
- When: Monday April 17, 2017, 11am-5pm (Reception following) (lunch will be served)
- Where: Technology Square Research Building (TSRB) Banquet Hall (map)
- RSVP required HERE (by April 3rd).
CyberLaunch an Atlanta-based accelerator for information security and machine learning startups hosting demo day on March 29, 2017 at 3pm
CyberLaunch is hosting it’s Winter 17′ Demo Day to showcase the current portfolio companies on March 29. The event will be held at the Atlanta Tech Village with over 250 investors and innovators, including 20+ cybersecurity and machine learning startups from around the world.
- Please register for our demo day HERE
- Location: Atlanta Tech Village – 3423 Piedmont Road NE – Atlanta, Georgia 30305
- If you are a startup looking to present at the startup showcase, apply HERE
- Date/Time: Wednesday, March 29 3pm
- Four companies, that are part of the 2nd Batch of this accelerator will do demo, followed a showcase of 20 cybersecurity and machine learning companies)
- For more details, see http://www.cyberlaunch.vc/demoday
ML@GT Seminar by Manos Antonakakis (ECE) on “Using DNS & Machine Learning to Reason About Internet Abuse” at 12n on March 29, 2017, in Nano 1117/8
Speaker: Manos Antonakakis, School of Electrical and Computer Engineering (ECE), GA Tech
Location: Marcus Nanotechnology 1117-1118 (Map)
Date/Time: March 29, 2017, 12:00n – 1:00pm (Lunch at 11:30am)Title: Using DNS & Machine Learning to Reason About Internet Abuse
Title: Using DNS & Machine Learning to Reason About Internet Abuse
Abstract: The Domain Name System (DNS) is a critical component of the Internet. The critical nature of DNS often makes it the target of direct cyber-attacks and other forms of abuse. Cyber-criminals rely heavily upon the reliability and scalability of the DNS protocol to serve as an agile platform for their illicit network operations. For example, modern malware and Internet fraud techniques rely upon the DNS to locate their remote command-and-control (C&C) servers through which new commands from the attacker are issued, serve as exfiltration points for the information stolen from the victim’s computer and to manage subsequent updates to their malicious toolset.
In this talk, I will discuss how we can reason about Internet abuse using DNS and various machine learning methods. After providing an overview around DNS, botnets and their illicit activities, I will discuss how spectral methods can help us model one of the most agile threats on the Internet; the botnets that employ Domain Name Generation Algorithms (DGAs). Then, we will discuss ways that tensors can help us track virtual illicit actors across the Internet. Finally, I will conclude by discussing some open research problems in computer security where machine learning methods should be the key ingredient for any efficient and effective solution.
Bio: Manos Antonakakis, Ph.D., is an Assistant Professor in the School of Electrical and Computer Engineering (ECE), and adjunct faculty in the College of Computing (CoC), at the Georgia Institute of Technology. He is responsible for the Astrolavos Lab, where students from both CoC and ECE conduct research in the areas of Network Security, Intrusion Detection, and Data Mining. In May 2012, he received his Ph.D. in Computer Science from the Georgia Institute of Technology. Before joining the ECE faculty, Professor Antonakakis held the Chief Scientist role at Damballa, where he was responsible for advanced research projects, university collaborations, and technology transfer efforts. He currently serves as the co-chair of the Academic Committee for the Messaging, Malware, and Mobile Anti-Abuse Working Group (M3AAWG). Dr. Antonakakis is the author of several U.S. patents and academic publications. He served as an external reviewer or a program committee member for leading information security conferences. He has successfully raised funding from multiple government agencies and organizations in the private sector. He is a member of the Institute for Information Security & Privacy (IISP) at Georgia Tech and contributed to its predecessor, the Georgia Tech Information Security Center (GTISC).
Ghassan Alregib’s Team (ECE) is hosting and running the IEEE VIP Cup in 2017 on the topic of “Traffic Sign Detection under Challenging Conditions”
Professor Ghassan Alregib’s team was been selected to be the team hosting and running the IEEE VIP Cup in 2017 on “Traffic Sign Detection under Challenging Conditions.” The topic is at the intersection of autonomous vehicles and machine learning. Not only it is attractive to the industry and the research communities but also the undergraduate students; the main goal is to get the UG students engaged with cutting-edge technologies. One of the side products is an open source dataset that has labeled ground truth data to test the vision algorithms under practical and real weather conditions, which have been missing in the publicly existing algorithms built for autonomous vehicles.
- The competition was announced today on the IEEE SPS website: https://signalprocessingsociety.org/get-involved/video-image-processing-cup.
- All the updates, the details, the datasets, and related info., are on the team website at https://ghassanalregib.com/vip-cup/.
Sixth Annual Workshop hosted by the Center for Energy and Geo Processing (CeGP)
The Center for Energy and Geo Processing (CeGP) hosted its 6th Annual workshop in Centergy (room 1010) on Monday and Tuesday, March 20-21, 2017. The workshop highlighted a number of talks from current and proposed CeGP projects that relate to advanced DSP and Machine Learning in Energy Applications. A few education presentations are also planned. CeGP is a partnership between Georgia Tech and University of Petroleum & Minerals (KFUPM) and is led by Dr. Ghassan AlRegib and Dr. Ali Al-Shaikhi
The details can be found at: http://cegp.ece.gatech.edu/events/workshop2017/.
NVIDIA and GT GPU Deep Learning Symposium, April 18
NVIDIA and GT GPU Deep Learning Symposium, April 18
Thanks to support from NVIDIA, we are hosting a deep learning symposium focused on the use of GPUs on April 18th, 10-3 in the Bill Moore Student Success Center. NVIDIA will be presenting some of its recent work with respect to deep learning as well as resources for students and researchers to accelerate their own work. In addition, several Georgia Tech and GTRI faculty are scheduled to speak about their machine learning work as it relates to GPU usage, including Le Song, Jimeng Sun, Zsolt Kira, and Oded Green.
If you are interested in attending please RSVP via the following link as space is limited. Please contact the event organizer, Jeff Young, if you have any questions.