Taylor T. Johnson, Associate Professor & Preston Robinette, PhD Student, Computer Science Vanderbilt University

Abstract: The ongoing renaissance in artificial intelligence (AI) has led to the advent of data-driven machine learning (ML) methods deployed within components for sensing, perception, actuation, and control in safety-critical cyber-physical systems (CPS). While such learning-enabled components (LECs) are enabling autonomy in systems such as autonomous vehicles and robots, ensuring such components operate reliably in all scenarios is extraordinarily challenging, as demonstrated in part through recent accidents in semi-autonomous/autonomous CPS and by adversarial ML attacks. In the first part of the talk, we will discuss formal methods for assuring specifications---mostly robustness and safety---in autonomous CPS and subcomponents thereof using our software tools NNV (https://github.com/verivital/nnv) and Veritex (https://github.com/verivital/veritex), developed partly in DARPA Assured Autonomy and Assured Neuro Symbolic Learning and Reasoning (ANSR) projects. In the second part of the talk, we will discuss recent results in analyzing neural networks used for malware classification through an ongoing National Security Agency (NSA) project. We will also discuss relevant ongoing community activities we help organize, such as the Verification of Neural Networks Competition (VNN-COMP) held with the International Conference on Computer-Aided Verification (CAV) the past few years, as well as the AI and Neural Network Control Systems (AINNCS) category of the hybrid systems verification competition (ARCH-COMP) also held the past few years. We will conclude with ongoing and future research directions in the broader safe and trustworthy AI domain, such as in a new project verifying neural networks used in medical imaging analysis.

Bios: Dr. Taylor T. Johnson, PE, is A. James and Alice B. Clark Foundation Chancellor Faculty Fellow and an Associate Professor of Computer Science (CS) in the School of Engineering (VUSE) at Vanderbilt University, where he directs the Verification and Validation for Intelligent and Trustworthy Autonomy Laboratory (VeriVITAL) and is a Senior Research Scientist in the Institute for Software Integrated Systems (ISIS). Dr. Johnson's research has been published in venues such as AAAI, CAV, EMSOFT, FM, FORMATS, HSCC, ICSE, ICDM, ICCPS, NFM, RTSS, SEFM, STTT, TNNLS, UAI, among others. Dr. Johnson earned a PhD in Electrical and Computer Engineering (ECE) from the University of Illinois at Urbana-Champaign in 2013, where he worked in the Coordinated Science Laboratory with Prof. Sayan Mitra, and earlier earned an MSc in ECE at Illinois in 2010 and a BSEE from Rice University in 2008. Dr. Johnson is a 2022 recipient of the Best Artifact Evaluation Award at FORMATS, a 2018 and 2016 recipient of the Air Force Office of Scientific Research (AFOSR) Young Investigator Program (YIP) award, a 2016 recipient of the ACM Best Software Repeatability Award at HSCC, a 2015 recipient of the National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) Research Initiation Initiative (CRII), and his group's research is or has recently been supported by AFOSR, ARO, AFRL, DARPA, Mathworks, NSA, NSF, NVIDIA, ONR, Toyota, and USDOT.

Preston Robinette is a 4th year PhD student in Computer Science at Vanderbilt University, and is a National Defense Science and Engineering Graduate Fellowship (NDSEG) receipient. Preston has done research and internships at Google, Apple, the National Security Agency (NSA), the Air Force Research Laboratory (AFRL), NASA Langley Research Center, the NASA South Carolina Space Grant Consortium, and Oak Ridge National Laboratory (ORNL). Preston's dissertation focus is at the intersection of machine learning and security, particularly deep learning approaches for steganography and watermarking.

Dr. Mary Hayhoe
Director of the Center for Perceptual Systems, University of Texas Austin

Abstract: A defining feature of the human visual system is that visual computations are task specific and linked to the location of gaze. This allows considerable computational efficiency. It also introduces the problem of what controls these computations. We think of this as the control of attention and gaze. In standard experimental paradigms, the experimental structure controls where human subjects attend. This is very different from vision in the natural world, where the subject determines the locus of attention. I will describe what is known about allocation of attention and control of gaze in natural visually-guided behavior. It has become clear that natural behavior has many regularities. We can think of the goal of visual processing as selecting good actions that further the organism’s behavioral goals. These decisions are modulated by the neural reward machinery of the brain, are made in the context of an uncertain and time-varying world, and involve both planning and memory. These elements of sequential decisions can be well described by Partially Observable Markov Decision Processes and there have been recent attempts to model natural gaze sequences in this way. I will summarize some of these attempts.

Bio: Mary Hayhoe is Director of the Center for Perceptual Systems at the University of Texas Austin. She received her Ph D from the University of California at San Diego and was a member of the Center for Visual Sciences at the University of Rochester from 1984-2006, when she moved to UT. She been a leader in developing virtual environments and experimental paradigms for the investigation of natural visually guided behavior. She has expertise in human eye movements in natural environments, especially how gaze behavior relates to attention, working memory, and cognitive goals. She served on the Board of Directors of the Vision Sciences Society from 2011-2016 and was President in 2015. She is also on the Editorial Board of the Journal of Vision. She was the recipient of the Davida Teller Award from the Vision Sciences Society in 2017 and the Tillyer Medal from Optica (formerly Optical Society of America) in 2022.