Indian Symposium on Machine Learning (IndoML)
December 16 – 18, 2020 | Virtual
The first Indian Symposium on Machine Learning (IndoML) will be hosted by the Indian Institute of Technology Gandhinagar (IITGN) between 16-18 December 2020. The symposium aims to be a forum to discuss state-of-the-art ML research through invited talks from leading experts within India and abroad. IndoML fosters mentoring of Indian Ph.D./Master students to network with their peers, seek expert guidance and develop early-stage collaborations.
IndoML aims to provide an opportunity for the faculty to engage with leading research groups in the country and conduct high-quality research leading to competitive publications. It will also provide a platform for industrial partners, including startups, working in ML-related areas to showcase their products and receive reviews/feedback as well as setup potential collaborations.
Theme

The theme of IndoML 2020 is “AI for Science and Science for AI“
Organizers
-
IIT Gandhinagar:
Anirban Dasgupta, Mayank Singh, Udit Bhatia -
IIT Kharagpur:
Animesh Mukherjee, Niloy Ganguly -
Publicity Chair:
Raj Sharma
Speakers

AADITESHWAR SETH
Indian Institute of Technology Delhi

ABIR DE
Indian Institute of Technology Bombay

ANKIT AGRAWAL
Northwestern University
Schedule
Mentioned time is Indian Standard Time (GMT+5 hr 30 mins)
Time (IST: GMT+5:30) |
Talk |
Speaker |
Day 2 Session I: Social Sciences |
||
Session Chair: Animesh Mukherjee (IITKGP) |
||
09:00 – 09:45 |
Mona Sloane: AI, Society, and Inequality: Why We Need to Enhance Engineers’ Ability to Understand Social Impact |
|
09:50 – 10:35 |
Joyojeet Pal: Social Media Misinformation in India |
|
10:40 – 11:25 |
Ravi Kumar: Learning-Augmented Online Learning |
|
Day 2 Session II: Chemical and Atmospheric Sciences |
||
Session Chair: Niloy Ganguly (IITKGP) |
||
16:00 – 16:45 |
Brooks Paige: Bayesian Inference and Meta-learning for Molecular Property Prediction with Graph Neural Networks |
|
16:50 – 17:35 |
Jose Miguel Hernandez-Lobato: Advances in Molecular Design with Deep Generative Models |
|
17:40 – 18:25 |
Kristof Schutt: Unifying Machine Learning and Quantum Chemistry with Deep Neural Networks |
|
18:30 – 18:50 |
Abir De: NeVAE – A Deep Generative Model for Molecular Graphs |
|
Time (IST: GMT+5:30) |
Talk |
Speaker |
Day 3 Session I: Physical and Mathematical Sciences |
||
Session Chair: Anirban Dasgupta (IITGn) |
||
08:15 – 09:00 |
Shirley Ho: Discovering Symbolic Models in Physical Systems using Deep Learning |
|
09:00 – 09:45 |
Anuj Karpatne: Science-guided Machine Learning – Advances in An Emerging Paradigm Combining Scientific Knowledge with Machine Learning |
|
09:50 – 10:35 |
J. Nathan Kutz: Machine Learning for Science: Data-Driven Discovery Methods for Governing Equations, Coordinates and Sensors |
|
Day 3 Session II: Social Sciences |
||
Session Chair: Mainack Mondal (IITKGP) |
||
16:00 – 16:45 |
Aaditeshwar Seth: Use of AI/ML to Scale Voice-based Participatory Media Forums |
|
16:50 – 17:35 |
Sudheendra Hangal: Challenging problems in Indian Political Data |
|
17:40 – 18:25 |
Milind Tambe: Advancing AI for Social Impact – Learning and Planning in the Data-to-Deployment Pipeline |
|
Registration
Registration is free of cost. The final registration confirmation would be based on the discretion of the organizers.
Sponsor

This event is being sponsored by a grant from the AI Journal.

AADITESHWAR SETH
Indian Institute of Technology Delhi
Title: Use of AI/ML to Scale Voice-based Participatory Media Forums
Abstract: Voice-based forums for discussion among local communities have had strong success in rural and less-literate communities, enabling them to share information and knowledge, local news, discuss policy, and demand accountability in local governance. Gram Vaani is a social enterprise that has been operating such voice forums in India since many years. I will talk about several interesting AI/ML problems that Gram Vaani is trying to solve to improve operational efficiency in handling large amounts of voice content.
Bio: Aaditeshwar Seth is a faculty at the Department of Computer Science and Engineering at IIT Delhi, and the co-founder and director of Gram Vaani, a social enterprise that uses technology to empower rural and low-income communities to run their own participatory media platforms.

Abir De
Indian Institute of Technology Bombay
Title: NeVAE: A Deep Generative Model for Molecular Graphs
Abstract: Deep generative models have been praised for their ability to learn smooth latent representation of images, text, and audio, which can then be used to generate new, plausible data. However, current generative models are unable to work with molecular graphs due to their unique characteristics-their underlying structure is not Euclidean or grid-like, they remain isomorphic under permutation of the nodes labels, and they come with a different number of nodes and edges. In this work, we first propose a novel variational autoencoder for molecular graphs, whose encoder and decoder are specially designed to account for the above properties by means of several technical innovations. Moreover, in contrast with the state of the art, our decoder is able to provide the spatial coordinates of the atoms of the molecules it generates. Then, we develop a gradient-based algorithm to optimize the decoder of our model so that it learns to generate molecules that maximize the value of certain property of interest and, given a molecule of interest, it is able to optimize the spatial configuration of its atoms for greater stability. Experiments reveal that our variational autoencoder can discover plausible, diverse and novel molecules more effectively than several state of the art models. Moreover, for several properties of interest, our optimized decoder is able to identify molecules with property values 121% higher than those identified by several state of the art methods based on Bayesian optimization and reinforcement learning.
Bio: Abir is an Assistant Professor at the Department of Computer Science and Engineering in IIT Bombay since January 2020. Before that, he was a postdoctoral researcher in Max Planck Institute for Software Systems at Kaiserslautern, Germany. He was hosted by Manuel Gomez Rodriguez. He received his PhD from Department of Computer Science and Engineering, IIT Kharagpur. His supervisor was Prof. Niloy Ganguly. During that time, he was a part of the Complex Network Research Group (CNeRG) at IIT Kharagpur. His PhD work was supported by Google India PhD Fellowship 2013 and got INAE best PhD thesis award. Prior to this he did his BTech in Electrical Engineering and MTech in Control Systems Engineering both from IIT Kharagpur.

Ankit Agrawal
Northwestern University
Title: AI and High Performance Data Mining – Illustrative Applications in Materials Science
Abstract: The increasing availability of data from the first three paradigms of science (experiments, theory, and simulations), along with advances in artificial intelligence and machine learning (AI/ML) techniques has offered unprecedented opportunities for data-driven science and discovery, which is the fourth paradigm of science. Within the arena of AI/ML, deep learning (DL) has emerged as a game-changing technique in recent years with its ability to effectively work on raw big data, bypassing the (otherwise crucial) manual feature engineering step traditionally required for building accurate ML models, thus enabling numerous real-world applications, such as autonomous driving. In this talk, I will present our ongoing research in AI and high performance data mining, along with illustrative real-world applications in materials science. In particular, we will discuss approaches to gainfully apply DL on big data (by accelerating DL and enabling deeper learning) as well as on small data (deep transfer learning) in the context of materials science. I will also demonstrate some of the materials informatics tools developed in our group.
Bio: Dr. Ankit Agrawal is a Research Professor in the Department of Electrical and Computer Engineering at Northwestern University, USA. He specializes in interdisciplinary AI and big data analytics via high performance data mining, based on a coherent integration of high performance computing and data mining to develop customized AI solutions for big data problems. His research has contributed to large-scale data-driven discoveries in various scientific and engineering disciplines, such as materials science, healthcare, social media, and bioinformatics. He has co-authored 150+ peer-reviewed publications, co-developed and released 15+ software, delivered 4 keynote and 40+ invited talks at major conferences, universities, and companies all over the world, been on program committees of 35+ conferences/workshops, and served as a PI/Co-PI on 15+ sponsored projects funded by various US federal agencies (e.g. NSF, DOE, AFOSR, NIST, DARPA, DLA) as well as industry (e.g. Toyota Motor Corporation Japan). In particular, he is one of the very few computer scientists who are actively introducing AI and advanced data science techniques in the field of materials science, and has successfully led several large-scale materials informatics projects. As an example, he is co-leading the AI group at the Center of Hierarchical Materials Design (CHiMaD), which is a $60-million NIST-sponsored center of excellence. He is also serving as the Editor-in-Chief of Computers, Materials & Continua.

Anuj Karpatne
Virginia Tech
Title: Science-guided Machine Learning: Advances in An Emerging Paradigm Combining Scientific Knowledge with Machine Learning
Abstract: This talk will introduce science-guided machine learning, an emerging paradigm of research that aims to principally integrate the knowledge of scientific processes in machine learning frameworks to produce generalizable and physically consistent solutions even with limited training data. This talk will describe several ways in which scientific knowledge can be combined with machine learning methods using case studies of on-going research in various disciplines including hydrology, fluid dynamics, quantum science, and biology. These case studies will illustrate multiple research themes in science-guided machine learning, ranging from physics-guided design and learning of neural networks to construction of hybrid-physics-data models. The talk will conclude with a discussion of future prospects in the emerging field of science-guided machine learning that has the potential to impact several disciplines in science and engineering that have a rich wealth of scientific knowledge and some availability of data.
Bio: Anuj Karpatne is an Assistant Professor in the Department at Computer Science at Virginia Tech and his primary interest is in developing data-enabled solutions for scientific and socially relevant problems. His current research is geared towards shaping the emerging field of research in Science-guided Machine Learning (SGML) (also referred to as physics-guided ML or theory-guided data science), where machine learning methods are systematically coupled with scientific knowledge (or physics) to accelerate scientific discovery. He enjoys working on inter-disciplinary projects and his research group is actively developing novel SGML formulations in a diverse range of scientific applications including lake modeling, remote sensing, fluid dynamics, quantum science, optics, radar physics, mechano-biology, and ichthyology.

Ard Louis
University of Oxford
Title: A Function Based Picture Explains Why Neural Networks Generalise So Well In The Overparameterized Regime
Abstract: One of the most surprising properties of deep neural networks (DNNs) is that they perform best in the overparameterized regime. We are all taught in a basic statistics class that having more parameters than data points is a terrible idea. This intuition can be formalised in standard statistical learning theory approaches, based for example on model capacity, which also predict that DNNs should heavily over-fit in this regime, and therefore not generalise at all. So why do DNNs work so well in a regime where theory says they should fail? A popular strategy in the literature has been to look for some dynamical property of stochastic gradient descent (SGD) acting on a non-convex loss-landscape in order to explain the bias towards functions with good generalisation. Here I will present a different argument, namely that DNNs are implicitly biased towards simple (low Kolmogorov complexity) solutions at initialisation [1]. This Occam’s razor like effect fundamentally arises from a version of the coding theorem of algorithmic information theory, applied to input-output maps [2]. We also show that for DNNs in the chaotic regime, the bias can be tuned away, and the good generalisation disappears. For highly biased loss-landscapes, SGD converges to functions with a probability that can, to first order, be approximated by the probability at initialisation [3]. Thus, even though, to second order, tweaking optimisation hyperparameters can improve performance, SGD itself does not explain why DNNs generalise well in the overparameterized regime. Instead it is the intrinsic bias towards simple (low Kolmogorov complexity) functions that explains why they do not overfit. Finally, this function based picture allows us to derive rigorous marginal-likelihood PAC-Bayes bounds that closely track DNN learning curves and that can be used to rationalise differences in performance across architectures.
[1] Deep learning generalizes because the parameter-function map is biased towards simple functions, Guillermo Valle Pérez, Chico Q. Camargo, Ard A. Louis arxiv:1805.08522
[2] Input–output maps are strongly biased towards simple outputs, K. Dingle, C. Q. Camargo and A. A. Louis Nature Comm. 9, 761 (2018)
[3] Is SGD a Bayesian sampler? Well, almost, Chris Mingard, Guillermo Valle-Pérez, Joar Skalse, Ard A. Louis arxiv:2006.15191
Bio: Ard Louis is professor of Theoretical Physics at the University of Oxford. He works on problems on a wide range of interdisciplinary problems.
https://www-thphys.physics.ox.ac.uk/people/ArdLouis/

Auroop Ganguly
Northeastern University
Title: Is Artificial Intelligence the new electricity that will transform Earth Systems Sciences and Engineering?
Abstract: There have been claims that AI is more profound than the discovery of fire or the invention of electricity as well as counter claims that AI will eventually destroy humanity but is not yet as mature as traditional disciplines like civil engineering. This presentation will examine the opportunities and challenges that AI may bring to bear, especially when combined with process understanding, in earth system sciences and engineering, both to advance fundamental sciences or engineering principles and for translating to societal priorities and policy imperatives.
Bio: Auroop R. Ganguly is a Professor of Civil and Environmental Engineering, as well as an Affiliate Professor of both the Khoury College of Computer Science and the School of Public Policy and Urban Affairs, at Northeastern University in Boston, MA. His research and teaching interests intersect water sustainability and climate science, urban sustainability and infrastructural resilience, as well as machine learning and nonlinear dynamics. He is a Fellow of the American Society of Civil Engineers and a Senior Member of the Institute of Electrical and Electronics Engineers. Prior to Northeastern, he was employed with the US DOE’s Oak Ridge National Laboratory and before that, at Oracle Corporation. Ganguly has published in journals such as Nature and PNAS, besides other journals across disciplines in geosciences, engineering, physics and data science, written a textbook on Critical Infrastructures Resilience, and won best paper awards in top-tier AI conferences. He is a co-founder of the Boston-based climate analytics startup risQ where he remains the chief scientific adviser. Ganguly has a PhD from the Massachusetts Institute of Technology in Cambridge, MA, and a B. Tech. (Hons.) from the Indian Institute of Technology at Kharagpur, India.

Brooks Paige
University College London AI Centre
Title: Bayesian Inference and Meta-learning for Molecular Property Prediction with Graph Neural Networks
Abstract: In this talk I will present recent work which aims to improve the generalization ability and low-data performance of graph neural networks for molecular property prediction. While graph neural networks have become a standard tool for in machine learning for molecules, thanks to their expressive power, there are two distinct challenges in this domain. The first is distributional shift: unlike in many other settings, the test molecules we would like to make predictions for nearly always differ significantly from the data used for training. Second, due to costs of data acquisition, molecular machine learning often operates in a low-data regime. As a result, graph neural networks which are underspecified by the data may not generalize well to new structure or scaffolds, and they may overfit to small training sets. We aim to address these problems by considering Bayesian estimation of graph neural networks, which captures uncertainty in model parameters and improves predictive accuracy and calibration; and by meta-learning, in which prediction tasks with few labels can benefit from other tasks where larger datasets are available.
Bio: Brooks Paige is an Associate Professor in Machine Learning in the Department of Computer Science at University College London. His research focuses on interpretable machine learning, methods for scalable Bayesian inference, and applications in the sciences. He holds a D.Phil in Engineering Science from the University of Oxford, where he was supervised by Frank Wood; an M.A. in Statistics from Columbia University; and a B.A. in Mathematics from Amherst College.

Chin-Yew Lin
Microsoft Research Asia
Title: Towards Automatic Math Word Problem Solving
Abstract: Computer programs can complete many tasks much more effectively and efficiently than human beings, such as calculating the product of two large numbers, or finding all occurrences of a string in a long text. However, the performance of computers on many intelligent tasks is still low. For example, in a chatting scenario, computers often generate irrelevant or incorrect responses; and we can easily find amusing results in automatic machine translations; and it is still a very challenging task for state-of-the-art computer programs to solve even primary-school-level math word problems. The SigmaDolphin project at MSRA aims to build a computer intelligent system that can automatically solve math word problems. In this talk, I will summarize our findings in addressing the three major challenges of math word problem solving: dataset creation, math word problem understanding and math equation generation.
Bio: Dr. Lin is a Principal Research Manager of the Knowledge Computing group at Microsoft Research Asia. His research interests are knowledge computing, natural language processing, semantic search, text generation, question answering, and automatic summarization. His team focuses on two main research directions: (1) grounded semantic computing, developing semantic computing framework for real world applications and services including automatic acquisition of world knowledge, machine reading for semantic indexing, automatic understanding of user intents, table interpretation and problem solving and (2) automatic text generation from structured data, developing models and algorithms to acquire writing and speaking knowledge automatically from massive data. The goal is to enable rich and context-aware interactive client plus cloud computing applications powered by automatically distilled and curated data. He developed automatic evaluation technologies for summarization, QA, and MT. The ROUGE automatic summarization evaluation package has become the de facto standard in summarization evaluations. It is the official automatic evaluation package for Document Understanding Conference. Before joining Microsoft, he was a senior research scientist at the Information Sciences Institute at University of Southern California (USC/ISI) where he worked in the Natural Language Processing and Machine Translation group since 1997. He is the Secretary-Treasurer of SIGDAT, was the program co-chair of ACL 2012, was program co-chair of AAAI 2011 AI & the Web Special Track, and was program co-chair of NLPCC 2016.

J. Nathan Kutz
University of Washington
Title: Machine Learning for Science: Data-Driven Discovery Methods for Governing Equations, Coordinates and Sensors
Abstract: Machine learning and artificial intelligence algorithms are now being used to automate the discovery of governing physical equations and coordinate systems from measurement data alone. However, positing a universal physical law from data is challenging: (i) An appropriate coordinate system must also be advocated and (ii) simultaneously proposing an accompanying discrepancy model to account for the inevitable mismatch between theory and measurements must be considered. Using a combination of deep learning and sparse regression, specifically the sparse identification of nonlinear dynamics (SINDy) algorithm, we show how a robust mathematical infrastructure can be formulated for simultaneously learning physics models and their coordinate systems. This can be done with limited data and sensors. We demonstrate the methods on a diverse number of examples, showing how data can maximally be exploited for scientific and engineering applications. The work also highlights the fact that the naive application of ML/AI will generally be insufficient to extract universal physical laws without further modification.
Bio: Professor Kutz is the Robert Bolles and Yasuko Endo Professor of Applied Mathematics at the University of Washington. He was awarded the B.S. in Physics and Mathematics from the University of Washington in 1990 and the PhD in Applied Mathematics from Northwestern University in 1994. Following postdoctoral fellowships at the Institute for Mathematics and its Applications (University of Minnesota, 1994-1995) and Princeton University (1995-1997), he joined the faculty of applied mathematics and served as Chair from 2007-2015. He is also adjunct professor of Physics, Mechanical and Electrical Engineering.

José Miguel Hernández-Lobato
University of Cambridge
Title: Advances in molecular design with deep generative models
Abstract: In this talk, I will present two recent contributions by my research group in the area of molecular design with deep generative models. First, I will describe “weighted retraining”, a new method that periodically retrains the generative model to account for newly collected data and also weights data points according to their objective value. Weighted retraining can be easily implemented on top of existing methods, and is empirically shown to significantly improve their efficiency and performance. Second, I will describe a new deep generative model that directly outputs molecule synthesis graphs. This provides sensible inductive biases, ensuring that the model searches over the same chemical space that chemists would also have access to. Our approach is able to model chemical space well, producing a wide range of diverse molecules, and allows for unconstrained optimization of an inherently constrained problem: maximize chemical properties such that discovered molecules are synthesizable.
Bio: José Miguel is a University Lecturer (equivalent to US Assistant Professor) in Machine Learning at the Department of Engineering in the University of Cambridge, UK. Before this, he was a postdoctoral fellow in the Harvard Intelligent Probabilistic Systems group at Harvard University, working with Ryan Adams, and a postdoctoral research associate in the Machine Learning Group at the University of Cambridge (UK), working with Zoubin Ghahramani. Jose Miguel completed his Ph.D. and M.Phil. in Computer Science at the Computer Science Department in Universidad Autónoma de Madrid (Spain), where he also obtained a B.Sc. in Computer Science from this institution, with a special prize to the best academic record on graduation. José Miguel’s research focuses on probabilistic machine learning, with a particular interest in deep generative models, Bayesian optimization, approximate inference, Bayesian neural networks and applications of these methods to the problem of automatic molecular design.

JOYOJEET PAL
Microsoft Research India
Title: Social Media Misinformation in India
Abstract: This talk examines the topical trends and flow trajectories of misinformation on social media in India. Examining Twitter, YouTube, and debunked misinformation in two news events – COVID-19, and the online activism following the death of actor Sushant Singh Rajput, we consider the weaponization of misinformation networks and its impact on the information environment in India.
Bio: Joyojeet Pal is a Principal Researcher at Microsoft Research India, where his work focuses on social media and society. His recent work has covered topics around the use of social media in politics, labour participation, and misinformation.

KRISTOF SCHÜTT
TU Berlin
Title: Unifying Machine Learning and Quantum Chemistry with Deep Neural Networks
Abstract: Deep neural networks are emerging as a powerful tool in quantum chemistry and materials science, combining the accuracy of electronic structure methods with computational efficiency. Going beyond the simple prediction of chemical properties, neural network potentials can be applied to perform fast molecular dynamics simulations including solvent effects, model response properties and generate novel structures with desired properties. On several examples, I will demonstrate how this opens a clear path towards increased synergy of machine learning and quantum chemistry as well as designing workflows tightly integrated with experiment.
Bio: Kristof Schütt is a senior researcher at the Berlin Institute for the Foundations of Learning and Data (BIFOLD). He received his master’s degree in computer science in 2012 and his PhD at the machine learning group of Technische Universität Berlin in 2018. Until September 2020, he worked at the Audatic company developing neural networks for real-time speech enhancement. His research interests include interpretable neural networks, representation learning, generative models, and machine learning applications in quantum chemistry.

LINQING WEN
University of Western Australia
Title: Potential and Challenges of Machine Learning for Gravitational Wave Discovery
Abstract: Gravitational Wave (GW) is an exciting field for the application of machine learning techniques. In this talk I will give an overview of the emerging field of gravitational wave astronomy, introduce the on-going research work at the University of Western Australia on the online detection of GWs using the live-streaming LIGO-Virgo detector data, and discuss the potential and challenges of applying machine learning techniques for gravitational wave discovery.
Bio: Linqing Wen is a professor of physics at the University of Western Australia (UWA). She obtained her PhD degree in Astrophysics from MIT in 2001 and was a postdoc at Caltech and Max Planck Institute for Gravitational Wave Physics in Germany before joining the UWA in 2007. She is a 2009 Australian Research Council Future Fellow and is currently leading the gravitational wave astronomy group at the UWA. Her recent research interests include online gravitational wave detection, high performance computing including machine learning, and astrophysics using gravitational wave data.

MAX WELLING
University of Amsterdam, and Qualcomm
Title: Lessons from Physics for Deep Learning
Abstract: A number of powerful principles underlie much of modern physics, such as the behavior of variables and fields under symmetry transformations and the strange statistical laws of quantum mechanics. Can these principles also be used in deep learning? While this may look strange at first sight, we only need to realize that both physics and deep learning can be understood as information processing systems. In this talk I will explain how we can apply representation theory for both global as well as local (gauge) transformations to deep learning. In the second half I will explain how even the language of quantum mechanics can be applied to deep learning and might, with the advent of quantum computers, become a new powerful paradigm for deep learning.
Bio: Prof. Dr. Max Welling is a research chair in Machine Learning at the University of Amsterdam and a VP Technologies at Qualcomm. He has a secondary appointment as a fellow at the Canadian Institute for Advanced Research (CIFAR). Max Welling has served as associate editor in chief of IEEE TPAMI from 2011-2015. He serves on the board of the Neurips foundation since 2015 and has been program chair and general chair of Neurips in 2013 and 2014 respectively. He was also program chair of AISTATS in 2009 and ECCV in 2016 and general chair of MIDL 2018. He is a founding board member of ELLIS. Max Welling is recipient of the ECCV Koenderink Prize in 2010. He directs the Amsterdam Machine Learning Lab (AMLAB), and co-directs the Qualcomm-UvA deep learning lab (QUVA) and the Bosch-UvA Deep Learning lab (DELTA).

Milind Tambe
Harvard University, and Google Research India
Title: Advancing AI for Social Impact –
Learning and Planning in the Data-to-Deployment Pipeline
Abstract: With the maturing of AI and multiagent systems research, we have a tremendous opportunity to direct these advances towards addressing complex societal problems. I focus on the problems of public health and wildlife conservation, and address one key cross-cutting challenge: how to effectively deploy our limited intervention resources in these problem domains. I present our deployments from around the world as well as lessons learned that I hope are of use to researchers who are interested in AI for Social Impact. Achieving social impact in these domains often requires methodological advances. I will highlight key research advances in topics such as computational game theory, multi-armed bandits and influence maximization in social networks as well as in integrating machine learning with such advances in the data to deployment pipeline. In pushing this research agenda, our ultimate goal is to facilitate local communities and non-profits to directly benefit from advances in AI tools and techniques.
Bio: Milind Tambe is Gordon McKay Professor of Computer Science and Director of Center for Research in Computation and Society at Harvard University; concurrently, he is also Director “AI for Social Good” at Google Research India. He is a recipient of the IJCAI John McCarthy Award, ACM/SIGAI Autonomous Agents Research Award from AAMAS, AAAI Robert S Engelmore Memorial Lecture award, INFORMS Wagner prize, Rist Prize of the Military Operations Research Society, Columbus Fellowship Foundation Homeland security award, AAMAS influential paper award, best paper awards at conferences such as AAMAS, IJCAI, IVA, and meritorious commendations from agencies such as the US Coast Guard and the Los Angeles Airport. Prof. Tambe is a fellow of AAAI and ACM.

Mona Sloane
Institute for Public Knowledge, NYU
Title: AI, Society, and Inequality: Why We Need to Enhance Engineers’ Ability to Understand Social Impact
Abstract: AI has been hailed as the gateway to the “4th Industrial Revolution”, mobilizing vast resource and sparking aspirations worldwide, while also sparking wide-spread fears of the robot-take over. As such, AI is an idea that is trapped between utopian and dystopian narratives. This polarization prevents us from examining the real potentials and pitfalls of AI, with severe consequences: how AI disproportionately inflicts harm on already disadvantaged populations is strategically disguised. What is also disguised is how these communities have to labor in order to protect themselves. This talk critically examines this dynamic and explores the link between AI, society, and inequality to argue for novel approaches in engineering education that embrace understanding the social impact of technology.
Bio: Mona Sloane is a sociologist working on inequality in the context of AI design and policy. She frequently publishes and speaks about AI, ethics, equitability and policy in a global context. Mona is a Fellow with NYU’s Institute for Public Knowledge (IPK), where she convenes the ‘Co-Opting AI’ series and co-curates the ‘The Shift’ series. She also works with NYU Vice-Provost Charlton McIlwain on building NYU’s new Alliance for Public Interest Technology, is an Adjunct Professor at NYU’s Tandon School of Engineering, and is part of the inaugural cohort of the Future Imagination Collaboratory (FIC) Fellows at NYU’s Tisch School of the Arts. Mona is also affiliated with The GovLab in New York and with Public Books where she curates the Technology section. Her most recent project is ’Terra Incognita: Mapping NYC’s New Digital Public Spaces in the COVID-19 Outbreak’ which she leads as principal investigator. Mona holds a PhD from the London School of Economics and Political Science and has completed fellowships at the University of California, Berkeley, and at the University of Cape Town. Follow her on Twitter @mona_sloane.

Ravi Kumar
Title: Learning-Augmented Online Learning
Abstract: In this talk we study online algorithms that are aided by predictions. In particular, we consider online optimization, where the algorithm receives a ‘hint’ regarding the upcoming cost vector before choosing the action for that round; we show near-optimal regret bounds in terms of how good are the hints. We also discuss extensions where the algorithm has access to multiple, possibly conflicting, hints at each round and where the algorithm must also pay for switching actions between rounds.

SATADEEP BHATTACHARJEE
Indo Korea Science and Technolgy Center (IKST)
Title: Learning-Augmented Online Learning
Abstract: In this talk we study online algorithms that are aided by predictions. In particular, we consider online optimization, where the algorithm receives a ‘hint’ regarding the upcoming cost vector before choosing the action for that round; we show near-optimal regret bounds in terms of how good are the hints. We also discuss extensions where the algorithm has access to multiple, possibly conflicting, hints at each round and where the algorithm must also pay for switching actions between rounds.

Shirley Ho
Flatiron Institute, and Carnegie Mellon University
Title: Discovering Symbolic Models in Physical Systems using Deep Learning
Abstract: We develop a general approach to distill symbolic representations of a learned deep model by introducing strong inductive biases. We focus on Graph Neural Networks (GNNs). The technique works as follows: we first encourage sparse latent representations when we train a GNN in a supervised setting, then we apply symbolic regression to components of the learned model to extract explicit physical relations. We find the correct known equations, including force laws and Hamiltonians, can be extracted from the neural network. We then apply our method to a non-trivial cosmology example—a detailed dark matter simulation—and discover a new analytic formula that can predict the concentration of dark matter from the mass distribution of nearby cosmic structures. The symbolic expressions extracted from the GNN using our technique also generalized to out-of-distribution-data better than the GNN itself. Our approach offers alternative directions for interpreting neural networks and discovering novel physical principles from the representations they learn.
Bio: Shirley Ho is an American cosmologist and astrophysicist, currently at the Center for Computational Astrophysics (Flatiron Institute) in NYC and at the Carnegie Mellon University. She obtained her Bachelor of Arts at the University of California, Berkeley with highest honors in Computer Science and Physics in 2004, and her Ph.D. at Princeton University in 2008. She significantly contributed to the development of several fields, including cosmic microwave background, dark energy, dark matter, the spatial distribution of galaxies and quasars, Baryon Acoustic Oscillations, and applications of machine learning to cosmology. More recently, Shirley Ho has led her team on a series of papers on accelerating simulations using modern deep learning techniques. Shirley Ho won several prizes for her significant contributions to the fields of cosmology and astrophysics. The list includes Macronix Prize (2014), Carnegie Science Award (2015), and she was named Cooper-Siegel Development Chair at Carnegie Mellon University (2015).

Srinivas (Sai) Ravela
MIT, and Galaxy.AI
Title: Stochastic Physics designs of Informative Hybrid Learning Machines for Earth, Planet, Climate and Life
Abstract: In machine learning applications in the physical sciences, to use a colleague’s phrase, a tussle between “Deep Learning” and “Deep Thinking” emerges. Classical stochastic processes have a well-established framework for integrating data with physically-based models in a system dynamics and optimization cycle that may typically involve model reduction, uncertainty quantification, estimation, and adaptive observation. These closed-loop processes are being modified by learning components for solving inverse problems, uncertainty quantification, parameterizing the “missing” physics, or up- or down-scaling. Whether hybrid systems emerge as Normal equations or as processes in their own right, part-ML part-physics solutions are now highly sought to, somehow, do better than either source alone. However, some critical questions about the efficacy and stability of the hybrid systems remain as they contend with nonlinear, nonstationary, and “infinite-dimensional” real-world processes. An appropriate formalism and tractable methodology are sorely missing.
In this talk, starting from a two-point boundary value problem description of learning, we study the Fokker Planck equations that lead to an “ensemble approach” to Deep Learning. This sets the basis for tractably quantifying uncertainty and information gain, which is used to optimize further the dynamics of learning, including establishing effective control on the stochastic learning process. Using this formalism, we will show how filters, smoothers, including fixed-lag smoothers, can be effectively developed for incremental online learning that produces stable hybrid systems. We discuss the unifying aspects of this Informative Learning approach for quantifying information transfer efficacy, optimizing input parameter and structure selection, and how it serves as a useful alternative for sparse optimization. Finally, we apply this methodology to three problems: Dynamic Downscaling of a turbulent process, learning equations from data, and optimizing neural structure, in each case showing the proposed approach to be efficacious.
Bio: Sai Ravela directs the Earth Signals and Systems Group at MIT where he is a Principal Research Scientist in the Earth, Atmospheric, and Planetary Sciences Department. His general research interests are in non-linear dynamics and stochastic processes with particular interest in information and inference. Dr. Ravela ‘s research group conducts research in animal biometrics for conservation, cooperative autonomous observation of the earth using unmanned aircraft, dynamic data driven modeling for estimating risk from natural hazards in a changing climate, fluid and seismic imaging, and exoplanet detection. Dr. Ravela’s group conducts methodological research in inference, information, learning for stochastic system dynamics and optimization. Dr. Ravela is the winner of the 2016 MIT Infinite Kilometer award for exceptional research and mentorship. He is the co-founder of Windrisktech LLC (Est. 2005). He is also the Head of Research and Development at Galaxy.AI where vision and learning technologies are being developed for intelligent damage detection. Dr. Ravela studied stochastic systems as a postdoc, computer vision, robotics, learning and information retrieval as a graduate student, earning a doctorate in computer science in 2002. For more information, please visit: http://essg.mit.edu

SUDHEENDRA HANGAL
Ashoka University
Title: Challenging problems in Indian Political Data