All meetings are held on Thursdays during lunchtime. 

Date: January 9
Topic: “Quantification and Low Dimensional Representation of Material Internal Structure Using Spatial Correlations”
Presenters(s): Surya R. Kalidindi
Abstract:
Almost all materials that relate to advanced technologies exhibit a richness of hierarchical internal structures at multiple length scales (spanning from atomic to macroscale). Certain salient features of this structure control the performance characteristics of interest for a selected application. Although there is often some intuition about what these salient features might be, validated protocols do not yet exist for reliably identifying these features. Further, efficient computational protocols do not yet exist for tracking their evolution during the various unit processing/synthesis steps employed in the industrial manufacture of new products/devices. In fact, the optimization of the material structure resulting in improved performance of engineering components is often the main motivation behind all activities in the field of materials science and engineering. Despite its important role, a unified computational framework for the quantification of the material hierarchical structure does not exist currently.  In this talk, I will describe a rigorous theoretical framework developed by my research group for the stochastic quantification of the material structure at any selected length scale, utilizing spatial correlations as the central metrics. This framework combines the use of n-point spatial correlations (or n-point statistics) and principal component analyses to arrive at objective, low-dimensional, representations of material internal structure in establishing core materials knowledge systems.  
Bio:
After earning a Ph.D. in mechanical engineering from the Massachusetts Institute of Technology in 1992, Surya R. Kalidindi joined the Department of Materials Science and Engineering at Drexel University. In 2013, his research group moved to the George W. Woodruff School at the Georgia Institute of Technology, where he has courtesy appointments in the School of Materials Science and Engineering and in the School of Computational Science and Engineering. Kalidindi’s research efforts over the past two decades have made seminal contributions to the fields of crystal plasticity and microstructure design. His current research is directed at developing and validating new data science enabled workflows for the successful realization of accelerated and cost-effective development of enhanced performance materials for advanced technologies.

All meetings are held on Thursdays during lunchtime. 

Date: January 16
Topic: “Scalable, Interactive, and Comprehensible Tools for Data Analytics”
Presenters(s): Polo Chau

Abstract:

Massive datasets now arise in virtually all domains. Yet, making sense of these data remains a fundamental challenge. At the “Polo Club of Data Science,” we are innovating at the intersection of data mining and human-computer interaction (HCI) to combine the best from both worlds to create scalable, interactive tools for making sense of graphs with billions of nodes and edges.

I will briefly describe some of our latest work, both on-going and published, that aims to tame big data through scalable algorithms, interactive visualization, and comprehensible models (machine learning/data mining) that users can more easily understand and work with.

Bio:

Duen Horng (Polo) Chau is an assistant professor in Georgia Tech’s School of Computational Science and Engineering and an adjunct assistant professor in the School of Interactive Computing. Polo received his Ph.D. from the Machine Learning Department at Carnegie Mellon University (CMU) and his Master’s degree in Human-Computer Interaction (HCI) from CMU.

Chau solves large-scale, “real world” problems that impact society. His NetProbe auction fraud detection system was featured in The Wall Street Journal and on CNN. His patented Polonium malware detection technology (with Symantec) protects 120 million people worldwide.

Chau’s thesis work received an honorable mention in Carnegie Mellon's School of Computer Science Distinguished Dissertation Award competition. He is the only two-time Symantec fellow, and he received a Yahoo! Key Scientific Challenges Award. Chau contributes to the PEGASUS peta-scale graph mining collaboration, which won an Open Source Software World Challenge Silver Award. He is also an award-winning designer and created Carnegie Mellon’s most recently used ID card design.

All meetings are held on Thursdays during lunchtime. 

Date: January 23
Topic: "Interactive Visual Analytics for Text Data Analysis"
Presenter: Haesun Park

Abstract:

Many modern data sets can be represented in high dimensional vector spaces and have benefited from computational methods that utilize advanced techniques from numerical linear algebra and optimization. Visual analytics approaches have contributed greatly to data understanding and analysis due to utilization of both automated algorithms and human’s quick visual perception and interaction. However, visual analytics targeting high dimensional large-scale data such as a document collection has been challenging due to low dimensional screen space with limited pixels to represent data.

We present some of the key foundational methods for supervised dimension reduction such as linear discriminant analysis (LDA), dimension reduction and clustering/topic discovery by nonnegative matrix factorization (NMF), and visual spatial alignment for effective fusion and comparisons by orthogonal Procrustes. We demonstrate how these methods can effectively support interactive visual analytic tasks that involve large-scale document data sets in two of the visual analytics systems, UTOPIAN for interactive topic discovery and VisIRR for visual document retrieval and recommendation.

Bio:

Prior to joining the College of Computing's School of Computational Science and Engineering faculty in 2005, Haesun Park was on the faculty at the University of Minnesota, Twin Cities, from 1987 to 2005, and from 2003 to 2005, she also served as a program director for the Computing and Communication Foundations Division at the National Science Foundation in Arlington, Va. She is the director of the NSF/DHS Foundations of Data and Visual Analytics (FODAVA-Lead) initiative and executive director of the Center for Data Analytics (CDA) at Georgia Tech.

Park's current research interests include numerical algorithms, pattern recognition, bioinformatics, information retrieval, and data mining. She has published more than 100 research papers in these areas and has served on numerous conference committees and journal editorial boards. Currently, she is on the editorial board of BIT Numerical Mathematics, SIAM Journal on Matrix Analysis and Applications, and the International Journal of Bioinformatics Research and Applications. In 2008 and 2009, Park served as a conference co-chair for the SIAM International Conference on Data Mining, and in 2013, she was elected as a SIAM Fellow.

All meetings are held on Thursdays during lunchtime. 

Date: February 6
Topic: "Understanding Language Variation in Social Media"
Presenter: Jacob Eisenstein

Abstract:

An increasing amount of informal communication is conducted in written form, through computer-mediated channels such as social media. This offers new challenges and opportunities for large-scale text analysis. I argue that social media analytics should be guided by an understanding of how social media language differs from other written language, and why. I will describe a series of studies that document language variation across a number of different social variables. In some cases, social media writing tracks spoken language variation; in other cases, relatively novel "netspeak" terms like emoticons and abbreviations can also be strongly correlated with demographics and geography. Our recent work concerns language change over time, using a new dataset of hundreds of thousands of authors over nearly three years. Aggregating across thousands of words, we build a unified model of the geographic and demographic factors that drive the spread of words between cities.

This research is in collaboration with David Bamman, Brendan O'Connor, Tyler Schnoebelen, Noah A. Smith, Eric P. Xing, and Yi Yang.

Bio:

Jacob Eisenstein is an assistant professor in the School of Interactive Computing at Georgia Tech. He works on statistical natural language processing, focusing on social media analysis, discourse, and latent variable models. Eisenstein was a postdoctoral researcher at Carnegie Mellon and the University of Illinois. He completed his Ph.D. at MIT in 2008, winning the George M. Sprowls dissertation award.

All meetings are held on Thursdays during lunchtime. 

Date: February 20
Topic: “Predictive Analytics for Improving the Reliability and Sustainability of Engineering Systems”
Presenter: Nagi Gebraeel

Abstract: 

Many high-valued engineering assets used in the manufacturing and service sectors are today being monitored by hundreds and perhaps thousands of sensors. Typically, the goal is to monitor system performance and degradation for numerous purposes, one of which is the prevention of unexpected failures. This talk focuses on how to effectively utilize sensor data to predict future system degradation and remaining lifetime (aka. prognostics).

The talk will begin by introducing a basic prognostic framework and how it has been implemented (through a live streaming demo, if possible). Next, the talk will focus on some scalability challenges that arise once we start dealing with big (sensor) data, which will be followed by a moderately technical discussion about some of the recently developed state-of-the-art modeling techniques that have targeted a few facets of this problem. 

Bio:

Nagi Gebraeel is the Chandler Family Associate Professor in the School of Industrial and Systems Engineering at Georgia Tech. He received his M.S. and Ph.D. degrees in industrial engineering from Purdue University in 1998 and 2003, respectively. Gebraeel’s research interests focus on leveraging condition-based sensor data streams to improve the predictability of unexpected failures of engineering systems and on improving subsequent operational and logistical decision making processes. He is a member of the Institute of Operations Research and Management Science (INFORMS), the Institute of Industrial Engineers (IIE), and the American Institute of Aeronautics and Astronautics (AIAA).

All meetings are held on Thursdays during lunchtime. 

Date: February 27
Topic: “Time Series Data in Healthcare: From Mental Health to the ICU”
Presenter: Gari Clifford

Abstract:

The enormous volume of available data from patients in both the hospital and outpatient settings presents both exciting opportunities for healthcare and several key problems. These include:

1. Traditional time series analysis algorithms are tuned to be over-sensitive, which results in many false alarms, and the expectation that an expert will over-read each alarm. With patient-to-doctor ratios ranging from 100 to 50,000 to one, this paradigm will no longer address the issues.
2. Multiple sensors can record similar information, and so we must build trust metrics to identify the signals we can trust, or work out ways to combine the signals together in a robust manner.
3. Humans disagree on diagnoses and labels, even when the disease is well described. Inter- and intra-human bias and variance in diagnoses must be addressed, particularly if we are to build automated algorithms from the labeled data.
4. Data labeling of medical data is vast and most likely impossible to do so by hand. Low-cost, semi-supervised and unsupervised approaches to event labeling, state definition, and feature extraction are needed. Finding clinically acceptable approaches to address this problem may be key to developing appropriate predictive machine learning algorithms.

Bio:

Gari Clifford recently joined the Emory and Georgia Tech faculties as an associate professor of bioinformatics and biomedical engineering. Previously, he was an associate professor of biomedical engineering at Oxford University and the director of the Centre for Doctoral Training in Healthcare Innovation at Oxford’s Institute of Biomedical Engineering.

Clifford’s research group explores machine learning and signal processing to extract actionable information from medical data. In particular, the lab focuses on intensive care medicine, cardiovascular disease, circadian rhythm disorders, sleep, and mental health. This research is aligned with the concept of sustainable healthcare, and he has a particular interest in mHealth in resource-constrained settings and circadian rhythms.

Prior to joining the faculty at Oxford, Clifford was a principal research scientist at MIT, where he spent six years managing the engineering effort behind a multi-million dollar project to collect and analyze the world's largest public database of hospital data. Clifford also serves on the international advisory and editorial boards of several organizations, including the NIH Public Access Resource, PhysioNet, and the Institute of Physics' Journal of Physiological Measurement.

In addition to licensing several patents, Clifford has been closely involved in the regulatory approval of medical devices for more than ten years. Also, he has received several awards for his research, including the 2009 Martin Black Prize, the 2010 mHealth Alliance Award, the 2011 International Engineering World Health Design Competition, the Dell Best Innovation Leveraging Technology Award in 2012, and the Computing in Cardiology Challenges in 2008, 2012, and 2013 for ECG analysis and mortality prediction.