Registration Check-in Desk closing at 6 pm. Badge pickup.
Given a machine learning model, what are the class of functions that can be approximated by this particular model efficiently, in the sense that the convergence rate for the approximation, estimation and optimization errors does not deteriorate as dimensionality goes up? We address this question for three classes of machine learning models: The random feature model, two-layer neural networks and the residual neural network model. During the process, we will also summarize the current status of the theoretical foundation of deep learning, and discuss some of the key open questions.
Can AI Be Used to Reduce Black Maternal Mortality?
Maternal mortality continues to be a global issue however, studies over the past 5 years identify a growing alarming trend in the United States with respect to maternal mortality. According to several studies from the Centers for Disease Control, “Black women are three times more likely to die from a pregnancy-related cause than White women”. Many of these same studies have also acknowledged that most pregnancy-related deaths are preventable and that this alarming trend continues to grow. This session will entail providing an overview of both past and current recommendations that are being used to address this issue which did not entail the usage of technology. Recent studies will also be identified that have indicated that Artificial Intelligence (AI) could be used to reduce Black maternal mortality. An exploration of how AI can be used to reduce Black maternal mortality and to advocate for further study of this topic will also be highlighted as a part of this session. During this session through conversation, dedicated reflection time, and creation of action plans we as a community will have an opportunity to explore what are ways that AI in conjunction with other resources can be used to address this ongoing and worsening issue.
Test of Time Award:
Poisoning Attacks Against Support Vector Machines
Battista Biggio, Blaine Nelson, Pavel Laskov:
Test of Time Honorable Mention:
Building high-level features using large scale unsupervised learning
Quoc Le, Marc'Aurelio Ranzato, Rajat Monga, Matthieu Devin, Kai Chen, Greg Corrado, Jeff Dean, Andrew Ng
On causal and anticausal learning
Bernhard Schölkopf, Dominik Janzing, Jonas Peters, Eleni Sgouritsa, Kun Zhang, Joris Mooij
Commonsense Knowledge Discovery and Acquisition
This social event will discuss current progress on commonsense knowledge acquisition and application with large models. Specifically, two invited speakers will be introducing their recent work on finding the knowledge bottleneck in pre-trained language models and commonsense knowledge discovery. After that, we will have a casual panel discussion to discuss some open questions regarding commonsense knowledge.
Trustworthy Machine Learning Social
A social to build a community for friends who are interested in trustworthy machine learning topics defined to the broadest scopes. Join us to meet the peers who might have read your papers or the peers you have read papers from, to know new friends who are interested in exploring topic in this dimension, and to build new collaboration opportunities