B. Aditya Prakash is an Associate Professor in the College of Computing at the Georgia Institute of Technology (“Georgia Tech”). He received a Ph.D. from the Computer Science Department at Carnegie Mellon University, and a B.Tech. in Computer Science and Engineering from the Indian Institute of Technology (IIT) -- Bombay. He has published one book, more than 90 papers in major venues, holds two U.S. patents and has given several tutorials at leading conferences. His work has also received multiple best-of-conference, best paper and travel awards. His research interests include Data Science, Machine Learning and AI, with emphasis on big-data problems in large real-world networks and time-series, with applications to computational epidemiology/public health, urban computing, security and the Web. Tools developed by his group have been in use in places like ORNL, CDC and Walmart. He has also received several awards such as the Facebook Faculty Award (2015 and 2021), the NSF CAREER award and was named as one of ‘AI Ten to Watch’ by IEEE. His work has won awards in multiple global data science challenges (e.g., the C3AI COVID19 Grand Challenge) and been highlighted by several media outlets/popular press like FiveThirtyEight.com too. He is a member of the infectious diseases modeling MIDAS network and core-faculty at the Center for Machine Learning (ML@GT) and the Institute for Data Engineering and Science (IDEaS) at Georgia Tech.

Aditya’s Twitter handle is @badityap.

In order to bring the performance of people analytics in the algorithmically infused workplace up — and in line with the hype — organizations need to do more than analyze data on demographic attributes. We need to focus not only on who people are but also who they know. The potential for social network analysis to identify “high potentials,” who has good ideas, who is influential, what teams will get work done efficiently and effectively is well established based on decades of research. The challenge has been the collection of network data via surveys that are time consuming, elicit low response rates and have a high obsolescence. This talk presents empirical examples ranging from corporate enterprises to simulated long duration space exploration to demonstrate how we can leverage people analytics – and in particular relational analytics - to mine “digital exhaust”— data created by individuals every day in their digital transactions, such as e‐mails, chats, “likes,” “follows,” @mentions, and file collaboration— to address challenges they face with issues such as team conflict, team assembly, diversity and inclusion, succession planning, and post-merger integration.

Bio

Noshir Contractor is the Jane S. & William J. White Professor of Behavioral Sciences in the McCormick School of Engineering & Applied Science, the School of Communication and the Kellogg School of Management and Director of the Science of Networks in Communities (SONIC) Research Group at Northwestern University. He is also the President-Elect of the International Communication Association (ICA).

Professor Contractor has been at the forefront of three emerging interdisciplines: network science, computational social science and web science. He is investigating how social and knowledge networks form – and perform – in contexts including business, scientific communities, healthcare and space travel. His research has been funded continuously for over 25 years by the U.S. National Science Foundation with additional funding from the U.S. National Institutes of Health, NASA, DARPA, Army Research Laboratory and the Bill & Melinda Gates Foundation.

His book Theories of Communication Networks (co-authored with Peter Monge) received the 2003 Book of the Year award from the Organizational Communication Division of the National Communication Association and the 2021 International Communication Association’s Fellows Book Award. He is a Fellow of the American Association for the Advancement of Science (AAAS), the Association for Computing Machinery (ACM), and the International Communication Association (ICA). He also received the Distinguished Scholar Award from the National Communication Association and the Lifetime Service Award from the Organizational Communication & Information Systems Division of the Academy of Management. He was selected as the recipient of the 2022 Simmel Award from the International Network for Social Network Analysis. In 2018 he received the Distinguished Alumnus Award from the Indian Institute of Technology, Madras where he received a Bachelor’s in Electrical Engineering. He received his Ph.D. from the Annenberg School of Communication at the University of Southern California.

Gus is a Developer Advocate in the Google AI team. His main passion is making Machine Learning easier for developers. As of lately he's been working with TensorFlow Hub and helping make ML models more accessible to everyone

ML on graphs is a prominent, and rapidly developing research area in recent years. Notably, graph neural networks (GNNs) have been proposed as a means of learning representations which utilize graph topology and node features, and are optimized toward targeted tasks of interest. Several such tasks deal with the challenging problem of learning with scarce, or nonexistent labeled data, common in domains like online trust and integrity. In this talk, I'll overview two recent works on learning in label-free and label-scarce settings on graph data with GNNs. The first work, published at TNNLS'21, discusses the use of integrating traditional unsupervised and modern GNN-based approaches for graph anomaly detection in a unified framework, demonstrating marked improvements in learned representation quality for anomaly detection tasks over both traditional and modern independent baselines. The second work, published at AAAI'21, discusses an approach for data augmentation for GNNs which shows large improvements in semi-supervised node classification, particularly in label-scarce settings. Together, the two works paint a promising picture for practical use of graph ML methods in settings with limited labels, by incorporating useful priors and stretching available labeled data with augmentation strategies. I'll also very briefly mention a few other exciting works our group has been up to, grounded in practically minded GNN research.

Bio: Neil Shah is a Lead Research Scientist at Snap Inc, with interests spanning data mining, machine learning and computational social science, specifically in the contexts of graph-based modeling of user behavior and misbehavior. His work has resulted in 35+ conference and journal publications, in top venues such as KDD, WSDM, ICDM, WWW, CIKM, SDM, AAAI, TKDD and more, including several best-paper awards. He has also served as an organizer, chair and senior program committee member at a number of these. He has had previous research experiences at Lawrence Livermore National Laboratory, Microsoft Research, and Twitch.tv. He earned a PhD in Computer Science in 2017 from Carnegie Mellon University's Computer Science Department, funded partially by the NSF Graduate Research Fellowship.

We consider the problem of learning efficient and inductive graph convolutional networks for text classification with a large number of examples and features. Existing state-of-the-art graph embedding based methods such as predictive text embedding (PTE) and TextGCN have shortcomings in terms of predictive performance, scalability and inductive capability. To address these limitations, we propose a heterogeneous graph convolutional network (HeteGCN) modeling approach that unites the best aspects of PTE and TextGCN together. The main idea is to learn feature embeddings and derive document embeddings using a HeteGCN architecture with different graphs used across layers. We simplify TextGCN by dissecting into several HeteGCN models which (a) helps to study the usefulness of individual models and (b) offers flexibility in fusing learned embeddings from different models. In effect, the number of model parameters is reduced significantly, enabling faster training and improving performance in small labeled training set scenario. Our detailed experimental studies demonstrate the efficacy of the proposed approach.

Bio: I did my Bachelors in Computer Science and Engineering at University of Mumbai. I went on to do my Masters at Indian Institute of Technology, Kanpur (IIT Kanpur) where I did my thesis in Reinforcement Learning. To get some experience in building practical AI/ML system, I spent 4 years as a Research Engineer at Yahoo! Research, Bangalore. There I worked on building large scale information tagging, extraction and management systems. Then I went on to do my PhD at Indian Institute of Technology, Bombay (IIT Bombay) under Prof. Sunita Sarawagi and Prof. Saketa Nath. During my PhD, I worked on building machine learning models for predicting aggregate label statistics. Currently, post my PhD, I have joined Microsoft Research, exploring two areas - 1] Learning Representations on Graphs and 2] Intersection of Program Synthesis and Machine Learning.

We investigate the capability of a transformer pretrained on natural language to generalize to other modalities with minimal finetuning -- in particular, without finetuning of the self-attention and feedforward layers of the residual blocks. We consider such a model, which we call a Frozen Pretrained Transformer (FPT), and study finetuning it on a variety of sequence classification tasks spanning numerical computation, vision, and protein fold prediction. In contrast to prior works which investigate finetuning on the same modality as the pretraining dataset, we show that pretraining on natural language can improve performance and compute efficiency on non-language downstream tasks. Additionally, we perform an analysis of the architecture, comparing the performance of a random initialized transformer to a random LSTM. Combining the two insights, we find language-pretrained transformers can obtain strong performance on a variety of non-language tasks.

Bio: Aditya Grover is a research scientist in the Core ML team at Facebook AI Research, a visiting postdoctoral researcher at UC Berkeley, and an incoming assistant professor of computer science at UCLA. His research centers around probabilistic modeling, approximate inference, and sequential decision making, with applications at the intersection of physical sciences and sustainability. His research has been recognized with a best paper award, a best undergraduate thesis award, several research fellowships (Microsoft Research, Lieberman, Google-Simons-Berkeley), and the ACM SIGKDD Doctoral Dissertation Award. Aditya received his PhD and bachelors in computer science from Stanford and IIT Delhi respectively.

Online hate speech is an important issue that breaks the cohesiveness of online social communities and even raises public safety concerns in our societies. Motivated by this rising issue, researchers have developed many traditional machine learning and deep learning methods to detect hate speech in online social platforms automatically. This talk aims to introduce the pressing problem of online hate speeches and discuss the automated hate speech detection methods.

Bio: Roy is an Assistant Professor at the Information Systems Technology and Design Pillar, Singapore University of Technology and Design. He is a faculty of the transformative Design and Artificial Intelligence programme. His research lies in the intersection of data mining, machine learning, social computing, and natural language processing. He is leading the Social AI Studio, a research group that focuses on designing the next-generation social artificial intelligence systems. He has published in top-tier venues in his relevant field, such as WWW, IJCAI, ICDM, TKDE, ACL, COLING, etc. He serves in the program committees of multiple top conferences. He is currently part of the editorial board for the Social Network Analysis and Mining (SNAM) journal.

In this talk I will present a set of experiments designed to help us understand what neural named entity recognition systems learn and why they make the predictions we see. Specifically, we seek to understand if systems learn name strings (Ani, Julia, Kathy) or if they are able to identify textual contexts that constrain the semantics class of whatever word appears in that context ("My name is __ "). I will present evidence that the performance of neural methods is largely driven by their ability to recognize word tokens as belonging to certain semantic classes. In a study with people, we find that in many cases people are indeed able to identify constraining contexts and figure out the class only from the context in the sentence. People's recognition of constraining contexts aligns better with predictions from biLSTM-CRF models than BERT models. I will present compelling evidence that current models do not integrate contextual clues effectively. These results indicate that NER is a challenging yet practical domain for testing machine text comprehension abilities.

Bio: Ani Nenkova is a Principal Scientist at Adobe Research and associate professor of computer and information science at the University of Pennsylvania (on leave). Her work on summarization, discourse, multi-modal emotion prediction and information extraction in the biomedical domain has been recognized with best paper awards at SIGDIAL 2010, EMNLP-CoNLL 2012, AVEC 2012 and AMIA 2021. Ani was program chair for NAACL in 2016 and currently serves as editor-in-chief for the Transactions of the Association for Computational Linguistics (TACL).

While we have made great progress in natural language understanding, transferring the success from benchmark datasets to real applications has not always been smooth. Notably, models sometimes make mistakes that are confusing and unexpected to humans. In this talk, I will discuss shortcuts in NLP tasks and present our recent works on guarding against spurious correlations in natural language understanding tasks (e.g. textual entailment and paraphrase identification) from the perspectives of both robust learning algorithms and better data coverage. Motivated by the observation that our data often contains a small amount of "unbiased" examples that do not exhibit spurious correlations, we present new learning algorithms that better exploit these minority examples. On the other hand, we may want to directly augment such "unbiased" examples. While recent works along this line are promising, we show several pitfalls in the data augmentation approach.

Bio: He He is an assistant professor in the Center for Data Science and Courant Institute at New York University. Before joining NYU, she spent a year at Amazon Web Services and was a postdoc at Stanford University. She received her PhD from University of Maryland, College Park. She is broadly interested in machine learning and natural language processing. Her current research interests include text generation, dialogue systems, and robust language understanding.

Current NLP pipelines rely significantly on finetuning large pre-trained language models. Relying on this paradigm makes such pipelines challenging to use in real-world settings since massive task-specific models are neither memory- nor inference-efficient, nor do we understand how they fare in adversarial settings. This talk will describe our attempts to address these seemingly unrelated concerns by investigating how specific short phrases in the input can control model behavior. These short phrases (which we call triggers) will help us identify model vulnerabilities and introduce new paradigms of training models. In the first part of the talk, I will focus on the adversarial setting. I will show how easy it is for adversaries to craft triggers that cause a target model to misbehave when the trigger appears in the input. I will also introduce a data poisoning technique that enables adversaries to inject arbitrary triggers into the target model. However, in the second part of the talk, I will show how these triggers can also be used to “prompt” language models to act as task-specific models, providing a negligible-memory, no-learning way to create classifiers. I will end with a comprehensive study of the interplay between prompting and finetuning, providing some guidelines for effectively performing few-shot learning with large language models.

Bio: Dr. Sameer Singh is an Associate Professor of Computer Science at the University of California, Irvine (UCI). He is working primarily on the robustness and interpretability of machine learning, and models that reason with text and structure for natural language processing. Sameer was a postdoctoral researcher at the University of Washington and received his Ph.D. from the University of Massachusetts, Amherst. He has received the NSF CAREER award, selected as a DARPA Riser, UCI ICS Mid-Career Excellence in research award, and the Hellman and the Noyce Faculty Fellowships. His group has received funding from Allen Institute for AI, Amazon, NSF, DARPA, Adobe Research, Hasso Plattner Institute, NEC, Base 11, and FICO. Sameer has published extensively at machine learning and natural language processing venues, including conference paper awards at KDD 2016, ACL 2018, EMNLP 2019, AKBC 2020, and ACL 2020.