Date: August 15, 2022, 8am-12pm EST

Room 102A

Scroll to the bottom for detailed agenda

Keynote Speakers

AutoDL: Research and Applications of Automated Deep Learning with Baidu’s Big Data

Big Data, Deep Learning, and huge computing are shaping up AI and are transforming our society. Especially, learning with deep neural networks has achieved great success across a wide variety of tasks. To help increase speed to solution and reduce duplication of effort, automated model construction is of great interest to provide architecture-effective and domain-adaptive deep models. On the other hand, understanding model behaviors and building trust in model prediction are essential for many applications such as autonomous driving, medical and fintech tasks. In this talk, we introduce Baidu AutoDL, which aims at building the next-generation ML system. Baidu AutoDL lowers the threshold for AI development and provides a full-stack deep learning solution, including (1) neural architecture search, (2) semi-supervised and transfer learning, (3) federated learning and (4) interpretable deep learning. State-of-the-art algorithms and industrial applications will be demonstrated in our talk.

Machine Learning in the Database

We have had databases, data mining, and business intelligence for decades, and machine learning for nearly the same number of years. So what's new and different now? There is now a much greater focus on bringing these specialized methods and algorithms to the typical business worker. This democratization of machine learning is essential since nearly every business worker is now digitally enabled, confronted with data every day, and expected to deliver insights and value from those data sources. The data sources (many and diverse), computing resources (Cloud), and algorithms of machine learning are sufficiently mature that bringing the benefits of these capabilities to every person in the business is important, imperative, and impactful. I will present some examples of one particularly significant recent development: machine learning in the database. Why this? Because nearly every business worker has access to and some familiarity with (if not, expertise in) database querying. Bringing anomaly detection, classification, predictive analytics (forecasting), or other machine learning capabilities into an SQL statement is a game-changer for enterprises everywhere. With ML in the database, democratized machine learning has never been so democratized!

Towards Understanding Users' Behaviors in Multi-Armed Bandits

Multi-Armed Bandits have proven to be an effective tool in many real-world applications, such as recommender systems, online advertising, and healthcare. In these applications, the ultimate goal is to satisfy the users' need from various aspects. Therefore, in this talk, I will introduce our recent work on modeling users' behaviors in multi-armed bandits. More specifically, I will start by introducing local user clustering in multi-armed bandits, which aims at leveraging user similarity to improve the quality of reward estimation; then I will present a new problem setting with multi-facet bandits, each characterizing the users' needs from one unique aspect. Towards the end, I will discuss the trade-off between exploitation and exploration in multi-armed bandits, and how we use a pair of deep models to learn the reward function as well as the potential gains.

Graph Neural Networks: Foundations, Frontiers, and Applications

The field of graph neural networks (GNNs) has seen rapid and incredible strides over the recent years. Graph neural networks, also known as deep learning on graphs, graph representation learning, or geometric deep learning, have become one of the fastest-growing research topics in machine learning, especially deep learning. This wave of research at the intersection of graph theory and deep learning has also influenced other fields of science, including recommendation systems, natural language processing, program synthesis, software mining, cybersecurity, and intelligent transportation. However, as the field rapidly grows, it has been extremely challenging to gain a global perspective of the developments of GNNs. Therefore, we feel the urgency to bridge the above gap and have a comprehensive book on this fast-growing yet challenging topic. In this talk, we will talk about our recent book titled "Graph Neural Networks: Foundation, Frontiers and Applications ", one of the most comprehensive books for researchers and practitioners for reading and studying in GNNs. It covers a broad range of topics in graph neural networks, by reviewing and introducing the fundamental concepts and algorithms of GNNs, new research frontiers of GNNs, and broad and emerging applications with GNNs.

Paper Presentations

• TIMEX: an Automatic Framework for Time-Series Forecasting-as-a-Service
  Alessandro Falcetta and Manuel Roveri

• ACE: Adaptive Constraint-aware Early Stopping in Hyperparameter Optimization
  Yi-Wei Chen, Chi Wang, Amin Saied and Rui Zhuang

• Mining Robust Default Configurations for Resource-constrained AutoML
  Moe Kayali and Chi Wang

• Lightweight Automated Feature Monitoring for Data Streams
  João Conde, Ricardo Moreira, João Torres, Pedro Cardoso, Hugo Ferreira, Marco Sampaio, João Tiago Ascensão and Pedro Bizarro

Agenda

8:00 – 8:40 AM Keynote 1: Dr. Dejing Dou - AutoDL: Research and Applications of Automated Deep Learning with Baidu’s Big Data

8:40 – 9:00 AM Paper 1: Mining Robust Default Configurations for Resource-constrained AutoML - Moe Kayali and Chi Wang

9:00 – 9:40 AM Keynote 2: Dr. Jingrui He - Towards Understanding Users' Behaviors in Multi-Armed Bandits

9:40 – 10:00 AM Paper 2: ACE: Adaptive Constraint-aware Early Stopping in Hyperparameter Optimization - Yi-Wei Chen, Chi Wang, Amin Saied and Rui Zhuang

10:00 – 10:40 AM Keynote 3: Dr. Lingfei Wu - Graph Neural Networks and structure discovery, with industry applications

10:40 – 11:00 AM Paper 3: Lightweight Automated Feature Monitoring for Data Streams - João Conde, Ricardo Moreira, João Torres, Pedro Cardoso, Hugo Ferreira,

Marco Sampaio, João Tiago Ascensão and Pedro Bizarro

11:00 – 11:40 AM Keynote 4: Dr. Kirk Borne - Machine Learning in the Database

11:40 – 12:00 PM Paper 4: TIMEX: an Automatic Framework for Time-Series Forecasting-as-a-Service - Alessandro Falcetta and Manuel Roveri