As is tradition for AusDM, we have lined up an excellent keynote speaker program. Each speaker is a well-known research and/or practitioner in data mining and related disciplines. The keynote program provides an opportunity to hear from some of the world’s leaders on what the technology oﬀers and where it is heading.
Talk 1: Why Multilayer Networks are Needed for Complex Data Analysis
Abstract: We are on the cusp of holistically analyzing a variety of diverse data being collected in every walk of life. For this, current analytics and science are being extended (Big Data Analytics/Science) along with new approaches. This warrants developing and/or using new approaches – technological, scientific, and systems – in addition to building upon and integrating with the ones that have been developed so far.
After going through the mining/analysis landscape, we present the need for multilayer networks for complex data analysis. Although mining/analysis of data using graphs has been around for a while, it has come to the forefront due to Internet and social networks as well as its versatility to model complex data sets. In this talk, we argue that graph analysis techniques are extremely important and hence renewed attention is needed as the data sets become complex.
We first illustrate the elegance of multilayer networks for modeling by using a few well-known data sets. We show how multiple entity and multiple relationships in data can be modeled elegantly using MLNs as compared to other representations. Flexibility of analysis comes from modeling the data MLNs. Then we discuss the “decoupling-based” approach for computation efficiency and scalability. This “divide and conquer” approach composes computations (e.g., communities, hubs, and substructures) from individual layers to form loss-less computation for any combination of layers in a multilayer network. Finally, we present several diverse case studies to showcase the approach.
Speaker: Professor Sharma Chakravarthy, The University of Texas at Arlington, Arlington, Texas
Prof. Sharma Chakravarthy is an ACM Distinguished Scientist and Distinguished speaker. He is also an IEEE Senior Member. He is also a Fulbright specialist. He organized (General Co-Chair) the 13th international Conference on Distributed Event-Based Systems (DEBS 2013). He has spent several summers at the Rome Air Force Research Laboratory (AFRL) as a Faculty Fellow working in continuous query processing over fault-tolerant networks and video stream analysis.
Sharma Chakravarthy is Professor of Computer and Engineering Department at The University of Texas at Arlington, Texas. He established the Information Technology Laboratory at UT Arlington in Jan 2000 and currently heads it. Sharma Chakravarthy has also established the NSF funded, Distributed and Parallel Computing Cluster (DPCC@UTA) at UT Arlington in 2003. He is the recipient of the university-level “Creative Outstanding Researcher” award for 2003 and the department level senior outstanding researcher award in 2002.
He is well known for his work on stream data processing, semantic query optimization, multiple query optimization, active databases (HiPAC project at CCA and Sentinel project at the University of Florida, Gainesville), and more recently scalability issues in graph mining, social network analysis, and graph analysis of multilayered networks. His group at UTA is currently adapting map/reduce and other paradigms for scaling graph mining algorithms to very large graphs and for answering graph queries. He has applied machine learning techniques to rank answers, identify general- and topic-based experts in a Question-Answer (or Q-A) social network. His work on InfoSift – a classification system for text, email, and web – has used graph mining techniques.
His current research includes big data analysis using multi-layered networks, stream data processing for disparate domains (e.g., video analysis), scaling graph mining algorithms for analyzing very large social and other networks, active and real-time databases, distributed and heterogeneous databases, query optimization (single, multiple, logic-based, and graph), and multi-media databases. He has published over 200 papers/book chapters in refereed international journals and conference proceedings. He has supervised 15+ PhD theses and 90+ MS thesis. He has given tutorial on a number of database topics, such as graph mining, active, real-time, distributed, object-oriented, and heterogeneous databases in North America, Europe, and Asia. He is listed in Who’s Who Among South Asian Americans and Who’s Who Among America’s Teachers.
Prior to joining UTA, he was with the University of Florida, Gainesville. Prior to that, he worked as a Computer Scientist at the Computer Corporation of America (CCA) and as a Member, Technical Staff at Xerox Advanced Information Technology, Cambridge, MA.
Sharma Chakrvarthy received the B.E. degree in Electrical Engineering from the Indian Institute of Science, Bangalore and M.Tech from IIT Bombay, India. He worked at TIFR (Tata Institute of Fundamental Research), Bombay, India for a few years. He received M.S. and Ph.D degrees from the University of Maryland in College park in 1981 and 1985, respectively.
Talk 2: An introduction to semi-supervised learning and contrastive loss
Abstract: In the last couple of years there has been a surge of interest in “self-supervised learning” (SSL). This is where we train a deep learning model using labels that are naturally part of the input data, rather than requiring separate external labels. SSL allows us to train with fewer labels, which has opened up many new domains to deep learning. Although the idea goes back to 1989 (Jürgen Schmidhuber), recent results have been particularly impressive thanks to the use of Large Language Models, and also the recent development of “consistency loss” (as it is known in NLP, or “noise contrastive estimation” in computer vision). This talk will cover the history of these ideas, the key ideas behind them, look at different approaches to implementing them, and discuss their results.
Speaker: Jeremy Howard, CSIRO, fast.ai
Jeremy Howard is a data scientist, researcher, developer, educator, and entrepreneur. Jeremy is a founding researcher at fast.ai, a research institute dedicated to making deep learning more accessible. He is also a Distinguished Research Scientist at the University of San Francisco, the chair of WAMRI, and is Chief Scientist at platform.ai.
Previously, Jeremy was the founding CEO Enlitic, which was the first company to apply deep learning to medicine, and was selected as one of the world’s top 50 smartest companies by MIT Tech Review two years running. He was the President and Chief Scientist of the data science platform Kaggle, where he was the top ranked participant in international machine learning competitions 2 years running. He was the founding CEO of two successful Australian startups (FastMail, and Optimal Decisions Group–purchased by Lexis-Nexis). Before that, he spent 8 years in management consulting, at McKinsey & Co, and AT Kearney. Jeremy has invested in, mentored, and advised many startups, and contributed to many open source projects.
He has many media appearances, including writing for the Guardian, USA Today, and the Washington Post, appearing on ABC (Good Morning America), MSNBC (Joy Reid), CNN, Fox News, BBC, and was a regular guest on Australia’s highest-rated breakfast news program. His talk on TED.com, “The wonderful and terrifying implications of computers that can learn”, has over 2.5 million views. He is a co-founder of the global Masks4All movement.