How to represent part-whole hierarchies in a neural network
Abstract : I will talk about my latest proposal for how to represent part-whole hierarchies in an neural network. I will illustrate the proposal with an imaginary system called GLOM. GLOM is a new way of doing shape recognition that solves many of the problems that capsules were designed to solve, but in a more elegant way. The talk will answer the question: How can a neural network with a fixed architecture parse images into part-whole hierarchies that are different for every image? GLOM suggests a way to significantly improve the interpretability of the representations produced by transformers when applied to language or vision.
Bio : Geoffrey Hinton received his PhD in Artificial Intelligence from Edinburgh in 1978. After five years as a faculty member at Carnegie-Mellon he became a fellow of the Canadian Institute for Advanced Research and moved to the Department of Computer Science at the University of Toronto where he is now an emeritus professor. He is also a VP Engineering fellow at Google and Chief Scientific Adviser at the Vector Institute. Geoffrey Hinton was one of the researchers who introduced the backpropagation algorithm and the first to use backpropagation for learning word embeddings. His other contributions to neural network research include Boltzmann machines, distributed representations, time-delay neural nets, mixtures of experts, variational learning and deep learning. His research group in Toronto made major breakthroughs in deep learning that revolutionized speech recognition and object classification. Geoffrey Hinton is a fellow of the UK Royal Society and a foreign member of the US National Academy of Engineering and the American Academy of Arts and Sciences. His awards include the David E. Rumelhart prize, the IJCAI award for research excellence, the Killam prize for Engineering, the IEEE Frank Rosenblatt medal, the NSERC Herzberg Gold Medal, the IEEE James Clerk Maxwell Gold medal, the NEC C&C award, the BBVA award, the Honda Prize and the Turing Award.
Demystifying Blockchains for Big Data
Abstract : Bitcoin is a successful and interesting example of a global scale peer-to-peer cryptocurrency that integrates many techniques and protocols from cryptography, distributed systems, and databases. The main underlying data structure is referred to as a blockchain, a scalable fully replicated structure that is shared among all participants and guarantees a consistent view of all user transactions by all participants in the system. In cryptocurrencies, the blockchain is often referred to as permissionless, since it operates in a setting where the identity, let alone the number, of participants is unknown. Furthermore, the participants are not trusted. Permissioned blockchains, on the other hand, assume the number of participants is known in advance, albeit possible that they are not trusted. The main application for permissioned blockchains are supply chain data infrastructures as well as other applications that require the notion of data provenance. In this talk, we will explore the opportunities and challenges in using blockchains for building large-scale decentralized data infrastructures in untrusted environments. We will also discuss how the atomic commitment paradigm from databases provides an excellent framework to support the fault-tolerant and decentralized atomic exchange of assets among different blockchains. This is particularly important given the proliferation of different crypto currencies where users need to atomically exchange assets without depending on centralized currency exchanges.
Bio : Divy Agrawal is a Professor of Computer Science at the University of California at Santa Barbara. His research interests are in the areas of databases, distributed systems, cloud computing, and big data infrastructures and analysis. He is the Fellow of the ACM, the IEEE, and the AAAS. He serves as the Editor-in-Chief of Journal of Distributed and Parallel Databases and has served on the Editorial boards of IEEE Transactions on Knowledge and Data Engineering, ACM Transactions on Database Systems, and ACM Transactions of Spatial Algorithms and Systems, and ACM Books. He has published 400+ articles on databases and distributed systems and has supervised 35+ PhD students during his tenure at the University of California at Santa Barbara. Dr. Agrawal is the recipient of the UCSB Academic Senate Award for Outstanding Graduate Mentoring.
Bringing Data Exchange to Knowledge Bases
Abstract : In data integration, theory and practice have long had a symbiotic relationship. This relationship is illustrated well in the history of data exchange tools and theory. In this talk, I discuss my experience in bringing data exchange to knowledge bases. This experience includes the development of Kensho, a tool for generating mapping rules and performing knowledge exchange between two Knowledge Bases (KBs). I highlight the challenges addressed in Kensho including handling the open-world assumption underlying KBs, managing the rich structural complexity of KBs, and the need to handle incomplete knowledge. I will use Kensho to highlight many open problems related to knowledge exchange including how knowledge translation can inform the task of KB integration and population.
Bio : Renée J. Miller is a University Distinguished Professor of Computer Science at Northeastern University. She is a Fellow of the Royal Society of Canada, Canada’s National Academy of Science, Engineering and the Humanities. She received the US Presidential Early Career Award for Scientists and Engineers (PECASE), the highest honor bestowed by the United States government on outstanding scientists and engineers beginning their careers. She received an NSF CAREER Award, the Ontario Premier’s Research Excellence Award, and an IBM Faculty Award. She formerly held the Bell Canada Chair of Information Systems at the University of Toronto and is a fellow of the ACM. Her work has focused on the long-standing open problem of data integration and has achieved the goal of building practical data integration systems. She and her colleagues received the 2013 ICDT Test-of-Time Award and the 2020 Alonzo Church Award for Outstanding Contributions to Logic and Computation for their influential work establishing the foundations of data exchange. Professor Miller is a former president of the Very Large Data Base (VLDB) Foundation and currently serves as Editor-in-Chief of the VLDB Journal. She received her PhD in Computer Science from the University of Wisconsin, Madison and bachelor’s degrees in Mathematics and Cognitive Science from MIT.
Systems are Predicates
Abstract : This talk is about Machine Learning at the intersection of Logic and Computer Science. Using system identification from biological data as a running thread, I intend to take you on a journey of using logic-based learning to identify simple mathematical functions, differential equation models,probabilistic transition systems and systems with feedback. Three fundamental ideas connect these results: (1) Systems are Programs (due to Gordon Plotkin); (2) Programs are Predicates (due to Tony Hoare); and (3) Predicates can be earned from data (the field of Inductive Logic Programming). If there is time, I will also touch on how logic-based learning can be combined usefully with neural-network learning, and in the construction of machine-assistants for scientific discovery.
Bio : Ashwin is the Class of 1981 Chair Professor at BITS Pilani, Goa. He received his Ph.D. in 1991 from the School of Electrical Engineering and Computer Science from the University of New South Wales. in 1991. Ashwin joined the ILP group at the Turing Institute, Scotland and---with S. Muggleton (now at Imperial College, London)--- worked on the development and application of Inductive Logic Programming.. From 1993--2003, Ashwin was at the Oxford University Computing Laboratory as a Research Fellow and then University Lecturer. At Oxford he was involved in pioneering applications of ILP systems to difficult real-world problems in molecular biology and chemistry. In 2003, he moved to the IBM Research – India, as a Research Staff Member. In 2009, he was awarded a Ramanujan Fellowship by the Department of Science and Technology of the Government of India. In 2010, he took up the post of Professor at the newly formed South Asian University, and became the founder Dean of the Faculty of Mathematics and Computer Science. He was Professor of Computer Science at the IIIT-D from 2012, and moved in Jan 2015 to BITS-Pilani, Goa as Chair Professor.
The Decision-Making Side of Machine Learning: Computational, Inferential and Economic Perspectives
Abstract : Much of the recent focus in machine learning has been on the pattern-recognition side of the field. I will focus instead on the decision-making side, where many fundamental challenges remain. Some are statistical in nature, including the challenges associated with multiple decision-making, and some are algorithmic, including the challenge of coordinated decision-making on distributed platforms. Finally, others are economic, involving learning systems that must cope with scarcity and competition. I will present recent progress on each of these fronts.
Bio : Michael I. Jordan is the Pehong Chen Distinguished Professor in the Department of Electrical Engineering and Computer Science and the Department of Statistics at the University of California, Berkeley. He received his Masters in Mathematics from Arizona State University, and earned his PhD in Cognitive Science in 1985 from the University of California, San Diego. He was a professor at MIT from 1988 to 1998. His research interests bridge the computational, statistical, cognitive and biological sciences. Prof. Jordan is a member of the National Academy of Sciences, a member of the National Academy of Engineering and a member of the American Academy of Arts and Sciences. He is a Fellow of the American Association for the Advancement of Science. He has been named a Neyman Lecturer and a Medallion Lecturer by the Institute of Mathematical Statistics. He was a Plenary Lecturer at the International Congress of Mathematicians in 2018. He received the Ulf Grenander Prize from the American Mathematical Society in 2021, the IEEE John von Neumann Medal in 2020, the IJCAI Research Excellence Award in 2016, the David E. Rumelhart Prize in 2015 and the ACM/AAAI Allen Newell Award in 2009. He is a Fellow of the AAAI, ACM, ASA, CSS, IEEE, IMS, ISBA and SIAM. In 2016, Professor Jordan was named the "most influential computer scientist" worldwide in an article in Science, based on rankings from the Semantic Scholar search engine.
Explaining the Explainable AI: A 2-Stage Approach
Abstract : Explainable AI (or XAI) has garnered a lot of interest in recent years across academia, industry and government. Although many methods have been proposed it is still unclear what XAI truly entails and why it is hard to formalize as opposed to other areas of machine learning such as causality, adversarial robustness, amongst others. In this talk, I will try to explicitly point out what XAI is trying to do, thus making it clear why formalization is difficult. The disentangled perspective also motivates new type of promising XAI approaches that are currently underexplored due to the largely intermingled view.
Bio : Amit Dhurandhar is Research Staff Member in the Trusted AI department at IBM T.J. Watson Research, where he leads the Explainable AI theme. His current research includes proposing various methods for enhancing trust in systems by developing methods that try to explain or understand their behaviors. His recent work was featured in Forbes and PC magazine with corresponding technical contribution in leading AI research venues such as NeurIPS, ICML and JMLR. His work has helped uncover interesting insights in fields such as Olfaction with papers in reputed journals such as Science and Nature Communications with extensive media coverage (Quartz, New Yorker, Atlantic, Biological Scene, Science News) . His research also has received the AAAI deployed application award as well as being selected as Best of IEEE ICDM. For his work on explainability he was invited to attend a Schloss Dagstuhl seminar in 2019. He was also one of the co-leads in the creation of the AI Explainability 360 open source toolkit. Besides research impact, his work has also gone into IBM products and his project work has received Outstanding Technical Achievement as well as a nomination for Corporate award. He has been an Area Chair and PC member for top AI conferences as well as has served on National Science Foundation (NSF) panels for the small business innovative research (SBIR) program.
Enabling convenience for the next billion through data science: A single beast or a swarm?
Abstract : When we look at what data science can do for enabling convenience at a high level, they all very nicely map to the same pattern of problems to be solved whether it is in mobility or hyperlocal delivery or e-commerce. When one gets deeper into solving these problems for all the variants, one starts realizing the nuances and complexities that make it very interesting for people looking at it through the lens of data science. I will be sharing what I learnt about the diversity and complexity swarm hiding behind what looks like a single beast via my experience at Ola, Swiggy and a bunch of other startups in the ecosystem.
Bio : Goda Ramkumar ranked as one of "Top 10 Data Scientists in India" by Analytics India Magazine in 2018, is currently Principal Data Scientist at Swiggy focusing on core logistics. Prior to this, she was GM, Data Science at xto10x establishing data science journey and practice in the start-up ecosystem. Previously she was at Ola for 2 years as Principal Data Scientist where she lead the machine learning and optimization algorithms behind pricing and matching. She worked at Sabre for 10 years before Ola on a variety of problems in Airline Pricing and Revenue Management. She has several publications and presentations in industry journals and academic conferences. She holds an undergraduate and masters degree from IIT Madras.
Frontiers in Natural Language Understanding for Conversational Platforms
Abstract : In this talk, I would like to share our experience in designing and improving the Natural Language Understanding (NLU) component of Siri. Siri is a popular conversational platform that enables seamless task-oriented interactions between humans and a variety of Apple devices, in multiple languages, processing billions of queries per-month worldwide. The talk will focus on discussing some of the recent problems in Siri NLU that demand deep-semantic understanding and how we addressed them through techniques like (a) dialog state tracking and (b) query rewriting. A portion of the talk will also be devoted to evaluating the NLU component, which is also key to assessing and improving Siri.
Bio : Abhijit Mishra is a Machine Learning Scientist, Apple inc., Seattle and drives some of the multilingual and cross-lingual efforts under the belt of Siri. His research works at Apple and IBM Research (prior to Apple) lie at the intersection of Multilingualism in NLP and Natural Language Generation. Abhijit is quite excited about semi-supervised data-to-text and text-to-text generation paradigms and believes in using them for (a) generating better response in dialog/interactive systems, and (b) preparing synthetic data for robust training of industrial NLP models. Some of his notable works include generating natural language descriptions from structured data such as knowledge graphs, unsupervised sarcasm generation, controllable style transfer, text simplification and template based multilingual data generation. Abhijit obtained Ph.D. from the Department of Computer Science and Engineering, IIT Bombay in 2017, under the guidance of Prof. Pushpak Bhattacharyya. His Ph.D. work explores a novel research direction, Cognitive NLP, that aims to- (i) uncover the cognitive underpinning of human language processing and (ii) translate the insights into better language processing systems. The thesis has been published as a monograph by Springer (https://link.springer.com/book/10.1007/978-981-13-1516-9). Since 2013, Abhijit’s works have been consistently getting published in the proceedings of prestigious NLP-AI conferences/journals such as Computational Linguistics, ACL, AAAI, WWW and EMNLP.
Topic: Challenges in deploying data science solutions
Panelists: Rajeev Rastogi (Amazon), Mayur Datar (Flipkart), Gargi Banerjee Dasgupta (IBM), Shailesh Kumar (Jio)
Moderator: Anand Deshpande (Persistent)
Bio : Mayur Datar works as a Chief Data Scientist with Flipkart in Bengaluru. He leads a large team of data scientists and together they are working on building the most advanced e-commerce landscape in India. Prior to joining Flipkart, Mayur worked for Google as a Research Scientist for over 12 years. At Google, Mayur led various projects namely Keyword Suggestion Tool for advertisers, Google Adwords Broad Matching, Google News personalization, Ratings/Reviews infrastructure, Adsense targeting for Social networks and others. He & his teams were credited with working on projects which had a big impact on Google’s bottom-line. Mayur has a doctorate in computer science from Stanford university and obtained his Bachelor of Technology from I.I.T. Bombay. He was awarded the President of India, Gold Medal for the ‘most outstanding student’ of his graduating batch at I.I.T. Bombay. He has several publications which have been presented in renowned conferences like SIGMOD, VLDB, KDD, FOCS, SODA, WWW. He serves on the review committees for most of these conferences. He is known in the industry for his technical leadership, pragmatic result oriented machine learning & systems design and experience in internet consumer related challenges. His research interests include data-mining, algorithms, databases and computer science theory.
Bio : Dr. Shailesh Kumar is currently the Chief Data Scientist at the Centre of Excellence in AI/ML, Reliance Jio. Prior to this he worked as a Distinguished Scientist at Ola cabs, Chief Scientist and Co-founder of Third Leap, an EdTech startup, Researcher in the Google Brain team, Sr. Scientist at Yahoo! Labs and Principal Scientist at Fair Isaac Research. Dr. Kumar has 18 years of experience in building AI solutions in a variety of domains including Web, Retail, Finance, Remote Sensing, Fleet Management, Computer Vision, Knowledge Graph, and Conversational computing. He has published over 20 international papers and book chapters and holds more than 20 patents in AI/ML. He was recognised as one of the top 10 data scientists in India in 2015 by Analytics India Magazine. Dr. Kumar holds a Masters and PhD in AI from UT-Austin and B.Tech. in Computer Science from IIT-Varanasi.
Bio : Gargi B. Dasgupta is the Director for IBM Research in India and the CTO for IBM India/South Asia. In her roles, Gargi is responsible for establishing and executing the technical agenda of IBM’s India Research Lab in collaboration with IBM’s worldwide research labs and business units. Gargi has been with IBM since 2004 when she joined as a Research Staff Member. Gargi is an Distinguished Engineer at IBM and was recognized as one of most powerful women in business by Fortune India in 2019, one of top women technologists by Economic Times 2020 and one of the most powerful women in business by Business Today 2020. Gargi is an ACM Distinguished Speaker and an alumnus of Jadavpur University, Kolkata and University of Maryland, Baltimore County.
Bio : Rajeev Rastogi is a Vice President of Machine Learning at Amazon where he is developing ML platforms and applications for the e-commerce domain. Previously, he was Vice President of Yahoo! Labs Bangalore and the founding Director of the Bell Labs Research Center in Bangalore, India. Rajeev is an ACM Fellow and a Bell Labs Fellow. He is active in the fields of databases, data mining, and networking, and has served on the program committees of several conferences in these areas. He currently serves on the editorial board of the CACM, and has been an Associate editor for IEEE Transactions on Knowledge and Data Engineering in the past. He has published over 125 papers, and holds over 100 patents. Rajeev received his B. Tech degree from IIT Bombay, and a PhD degree in Computer Science from the University of Texas, Austin.
Bio : Dr. Anand Deshpande is the Founder, Chairman and Managing Director of Persistent Systems since its inception and is responsible for the overall leadership, strategy and management of the Company. Anand holds a B. Tech. (Hons.) in Computer Science and Engineering from IIT Kharagpur, and a M.S. and Ph.D. in Computer Science from Indiana University, Bloomington, Indiana, USA. As a true technology visionary, Anand’s strengths lie in identifying and investing in next-generation technologies and encouraging internal entrepreneurship to ensure that Persistent Systems stays at the forefront of technology innovation. Anand has been the driving force in growing Persistent Systems from its inception in 1990, to the publicly-traded global Company of today. He has been recognized by his alma mater, IIT Kharagpur, as a Distinguished Alumnus in 2012 and by the School of Informatics of Indiana University with the Career Achievement Award in 2007. Prior to founding Persistent Systems, Anand began his professional career at Hewlett-Packard Laboratories in Palo Alto, California, where he worked as Member of Technical Staff from May 1989 to October 1990. Currently, he serves as a Trustee of the VLDB Endowment (www.vldb.org), Trustee of BAIF and as a Trustee of Persistent Foundation.
Topic: Challenges and Opportunities presented by COVID-19 for the Data Science community
Panelists: Harshad Khadilkar (TCS), Srujana Merugu (Google Research), Mitesh M. Khapra (IIT Madras)
Moderator: Karthik Sankaranarayanan (IBM)
Bio : Harshad Khadilkar is a scientist with the research division of Tata consultancy services, where he leads the Planning & Control team in Mumbai. He is also a visiting associate professor with the aerospace engineering department at IIT Bombay. He holds a bachelors degree from the same department, in addition to masters and PhD degrees from the Massachusetts Institute of Technology. His research interests are in control and reinforcement learning algorithms for networked systems.
Bio : Srujana is a machine learning researcher (currently consulting with Google research & Wadhwani AI) with a keen interest in platform dynamics, data governance, human behavior, education, healthcare, and financial inclusion. Prior to 2017, she was employed with the machine learning groups at Yahoo Research, IBM Research, Amazon, and Flipkart. Post-2017, she has mostly been involved in consulting engagements with startups in education, health-care, CRM, SaaS, and Fintech domains along with volunteering on social causes. Since March of this year, she has been assisting with the COVID response of a few highly impacted cities and states as part of the COVID-19 Data Science Consortium. She has published her research in several top-tier conferences and journals on machine learning and is the recipient of multiple best paper awards. She received her M.S. and Ph.D. from the University of Texas at Austin and her B. Tech. degree from IIT Madras.
Bio : Mitesh M. Khapra is an Assistant Professor in the Department of Computer Science and Engineering at IIT Madras and is affiliated with the Robert Bosch Centre for Data Science and AI. He is also a co-founder One Fourth Labs, a startup whose mission is to design and deliver affordable hands-on courses on AI and related topics. He is also a co-founder of AI4Bharat, a voluntary community with an aim to provide AI-based solutions to India-specific problems. His research interests span the areas of Deep Learning, Multimodal Multilingual Processing, Natural Language Generation, Dialog systems, Question Answering and Indic Language Processing. He has publications in several top conferences and journals including TACL, ACL, NeurIPS, TALLIP, EMNLP, EACL, AAAI, etc. He has also served as Area Chair or Senior PC member in top conferences such as ICLR and AAAI. Prior to IIT Madras, he was a Researcher at IBM Research India for four and a half years, where he worked on several interesting problems in the areas of Statistical Machine Translation, Cross Language Learning, Multimodal Learning, Argument Mining and Deep Learning. Prior to IBM, he completed his PhD and M.Tech from IIT Bombay in Jan 2012 and July 2008 respectively. His PhD thesis dealt with the important problem of reusing resources for multilingual computation. During his PhD he was a recipient of the IBM PhD Fellowship (2011) and the Microsoft Rising Star Award (2011). He is also a recipient of the Google Faculty Research Award (2018), the IITM Young Faculty Recognition Award (2019) and the Prof. B. Yegnanarayana Award for Excellence in Research and Teaching (2020).
Bio : Dr. Karthik Sankaranarayanan is a Senior Technical Staff Member (STSM) and Senior Manager in Artificial Intelligence at IBM Research, Bangalore where leads a team of scientists and engineers working at the intersection of deep learning, natural language processing and computer vision. At IBM, his research is highly influential in applications of AI towards next generation of challenges in Financial Services and Healthcare for IBM Watson. His work has received several technical accomplishments including IBM Corporate Award -- the highest honor at IBM awarded by IBM CEO. He also serves as the IBM Principal Investigator for the IBM - IIT Bombay AI Horizons Network (AIHN) strategic partnership program. Before joining IBM Research in 2011, he obtained his PhD in Computer Science in machine learning and computer vision from The Ohio State University, USA where he was a U.S. Dept. of Energy Graduate Fellow between 2009 and 2011. His work has been published extensively at flagship AI conferences and journals such as NeurIPS, ACL, AAAI, CVPR, Interspeech, KDD, VLDB. He hold more than 50 patents around applications of AI to industry problems.
PhdLife: Different Strokes for Different Folks
Abstract : If you are asking any or all of these questions, tune in to find some answers. I will share my experience of a decade in answering these kinds of questions. Will have some prescriptive and abstract ways by which your Ph.D. life can be made more exciting, productive, and perhaps, a life changing experience. How do I find out if research is something I like the most and want to spend time in doing a Ph.D. and beyond? I am scared of committing 5 years for a Ph.D., will it be worth the time? How can I find out if I will be able to finish my Ph.D. well? How to find a good thesis advisor? How is the Advisor-Advisee relationship? Many of my friends are joining / doing Ph.D., should I? What are the tools, techniques that successful Ph.D. students use? How important is networking, getting to know people in your area of work? How important is social media, and is there a way to effectively use it during Ph.D. life?.
Bio : Prof. Ponnurangam Kumaraguru ("PK") is a Professor of Computer Science and Dean of Students Affairs at IIIT-Delhi. He is a Visiting Faculty at IIT Kanpur and an Adjunct faculty at IIIT Hyderabad. PK is an ACM India Council Member, and Chair of the Publicity & Membership Committee of ACM India. PK is a TEDx and an ACM Distinguished & ACM India Eminent Speaker. PK received his Ph.D. from the School of Computer Science at Carnegie Mellon University (CMU). His Ph.D. thesis work on anti-phishing research at CMU contributed in creating an award-winning startup - Wombat Security Technologies, wombatsecurity.com. Wombat was acquired in March 2018 for USD 225 Million. In addition to his contributions to academia, PK is on advisory role on various government organizations and a Fortune 500 company. He is also part of various Board of Studies of different academic institutes across the country. He has co-authored research papers in the field of Privacy and Security in Online Social Media, Cyber Security, Computational Social Science, Social Computing, Data Science for Social Good, amongst others. He is passionate about solving societal problems using technologies, especially contributing to the domain of social computing. PK and his students have played an integral role in developing a technology used many State and Central Government agencies in India. He has spent time making notable contributions at organizations like IBM India Research Labs, Adobe Research Labs – India, Universidade Federal de Minas Gerais (UFMG) – Brazil, and Max Planck Institute for Software Systems – Germany. PK has served as a Program Committee member at prestigious conferences like WWW, CHI, PETs, ICWSM, CSCW, AsiaCCS. He was the Founding Head of Cybersecurity Education and Research Centre (CERC) at IIIT-Delhi. PK started and successfully manages PreCog (precog.iiitd.edu.in), a research group at IIIT-Delhi.