Week 12 Activity (Thursday): Colloquium talk at Mudd

This week, Harvey Mudd College hosts a talk by Andrew (Haoze) Wu, from Stanford University, who is a candidate for a faculty position in the CS department at the college. The talk begins at 4:15 PM on Thursday, but a reception with refreshments will be held outside at 4:00 PM.

There are two colloquium talks this week, so we have slightly different rules for this week for students enrolled in CS Colloquium (CS 195).

• We would be hugely pleased to have you attend both talks. These talks don't only let us see a faculty candidate, they also let the faculty candidate see us, so having a good audience is important. But if you can only manage to see a single talk this week, that is okay too.
• If you are in Section 1, we know you're free to attend this talk, but if you'd like to attend the other talk on Tuesday, feel free to do so, either as well or instead.
• If you are in Section 2, and you cannot attend Tuesday's talk, you can watch the recording of either talk.

AI Meets AR

Abstract

Methods from Artificial Intelligence (AI), such as Deep Learning, have the potential to profoundly benefit the world. However, due to the unstable and opaque nature of Deep Learning models, ensuring their safe and reliable deployment, especially in high-stakes applications, remains a crucial challenge. Fortunately, while uninterpretable to humans, Deep Learning models can be analyzed computationally, using formal reasoning and logic. Towards this end, I present an Automated Reasoning (AR) framework that can efficiently answer, with mathematical rigor, whether a Deep Learning model possesses a given property. I will show how this framework can be used to examine and obtain provable guarantees about the behaviors of models used in a wide range of applications, from unmanned aircraft control to data center job scheduling. By significantly improving the scalability and expressivity of Automated Reasoning techniques for AI, my work unlocks the potential to create provably safe and robust machine-learning-driven computer systems.

About Andrew (Haoze) Wu

Andrew (Haoze) Wu is a PhD candidate in Computer Science at Stanford University. He obtained his bachelor's degree in mathematics and philosophy from Davidson College. Andrew's work aims to bring together automated reasoning and machine learning to create safer and smarter computer systems. He is the lead developer of the Marabou framework, a state-of-the-art neural network verification tool widely used in academia and industry. His work has been published at top venues in different fields, including CAV, FMCAD, TACAS, OOPSLA, AISTATS, NeurIPS, and IROS.

When and How to Attend

• Thursday, November 16
  • Location: Galileo McAlister, Harvey Mudd College
  • Optional reception begins at 4:00 PM
  • Talk runs from 4:15–5:30 PM

Recording for Section 2

Required Assessment

To receive credit for attending this colloquium, complete the assessment:

• Assessment

Please do so at your soonest convenience, within 24 hours of seeing the talk.

(When logged in, completion status appears here.)

Previous Page Go to Parent Page