- Ai - Engineering Mailing Lists

AI Seminar : Jan 26
by Mangannavar, Rajesh Devaraddi 24 Jan '24

24 Jan '24

Dear all, Our next AI seminar is scheduled to be on January 26th, 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Value-Based Abstractions for Planning Amy Zhang Assistant Professor Chandra Family Department of Electrical and Computer Engineering University of Texas at Austin Abstract: As the field of robotics continues to advance, the integration of efficient planning algorithms with powerful representation learning becomes crucial for enabling robots to perform complex manipulation tasks. We address key challenges in planning, reward learning, and representation learning through the objective of learning value-based abstractions. We explore this idea via goal-conditioned reinforcement learning, action-free pre-training, and with language. By leveraging self-supervised reinforcement learning and efficient planning algorithms, these approaches collectively contribute to the advancement of robotic systems capable of learning and adapting to diverse tasks in real-world environments. Speaker Bio: I am an assistant professor at UT Austin in the Chandra Family Department of Electrical and Computer Engineering. My work focuses on improving generalization in reinforcement learning through bridging theory and practice in learning and utilizing structure in real world problems. Previously I was a research scientist at Meta AI - FAIR and a postdoctoral fellow at UC Berkeley. I obtained my PhD from McGill University and the Mila Institute, and also previously obtained an M.Eng. in EECS and dual B.Sci. degrees in Mathematics and EECS from MIT. Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 0

AI Semin : Jan 19
by Mangannavar, Rajesh Devaraddi 17 Jan '24

17 Jan '24

Dear all, Our next AI seminar is scheduled to be on January 19th, 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> AI for Software Engineering (AI4SE): Opportunities and Challenges Manish Motwani Assistant Professor School of Electrical Engineering and Computer Science Oregon State University Abstract: Software is ubiquitous but often faulty. Software engineers spend 35–50% of their time debugging software, accounting for 50–75% of the total software development budget. With the transformative power of AI, software engineering researchers have begun integrating AI into various phases of the software development lifecycle to address these issues and enhance software quality. In this talk, I will provide a high-level overview of opportunities and challenges for applying AI in software engineering, specifically focusing on software testing and debugging. I will then present my research on using natural language artifacts to automatically improve software quality, including my technique for automatically generating tests with oracles from software specifications and my method for using developer-written tests in conjunction with bug reports to automatically localize and repair bugs. Speaker Bio: Manish Motwani is an Assistant Professor in the School of Electrical Engineering and Computer Science at Oregon State University. Before joining OSU, he worked as a postdoc at Georgia Tech and has four years of industry work experience as a researcher at the Tata Research Development and Design Center, an industrial research lab in India. His interests lie in improving software engineers’ productivity and aiding software engineers in developing high-quality software by automating software engineering practices. His research involves learning phenomena from large software repositories and using that knowledge to design novel automation techniques, such as requirements elicitation, test synthesis, and program repair. Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 0

AI Seminar : Jan 12
by Mangannavar, Rajesh Devaraddi 10 Jan '24

10 Jan '24

Dear all, Our next AI seminar is scheduled to be on January 12th, 2-3 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Reinforcement Learning from a Bayesian perspective Brendan O'Donoghue Research Scientist DeepMind Abstract: Reinforcement learning (RL) involves an agent interacting with an environment over time attempting to maximize its total return. Initially the agent does not know about the environment and must learn about it from experience. As the agent navigates the environment it receives noisy observations which it can use to update its (posterior) beliefs about the environment. Therefore, the RL problem is a statistical inference problem wrapped in a control problem, and the two problems must be tackled simultaneously for good data efficiency. This is because the policy of the agent affects the data it will collect, which in turn affects the policy, and so on. This is in contrast to supervised learning, where the performance of a classifier (for instance) does not influence the data it will later observe. Failure to properly consider the statistical aspect of the RL problem will result in agents that require exponential amounts of experience for good performance. On the other hand, correctly considering the statistical inference problem and the control problem together has the potential to dramatically reduce the compute requirements to solve problems and potentially unlock new domains and capabilities far outside of the range of current agents. In this talk I will introduce these concepts and discuss how Bayesian techniques can provide principled solutions to the problem. Speaker Bio: Brendan O'Donoghue earned his PhD in 2013 from Stanford working with Stephen Boyd on optimization and control theory. Since then he has worked at DeepMind as a research scientist working on deep reinforcement learning, optimization, and (more recently) large language models. Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 0

AI Seminar: Brendan O'Donoghue
by Tadepalli, Prasad 10 Jan '24

10 Jan '24

Hi all, The first AI Seminar this term is by Brendan O'Donoghue of Deep Mind on Friday, 1/12 2-3 pm (new time) in KEC 1001, which will be followed by a 30 minutes time for Q&A. Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ… Title: Reinforcement Learning from a Bayesian Perspective Abstract Reinforcement learning (RL) involves an agent interacting with an environment over time attempting to maximize its total return. Initially the agent does not know about the environment and must learn about it from experience. As the agent navigates the environment it receives noisy observations which it can use to update its (posterior) beliefs about the environment. Therefore, the RL problem is a statistical inference problem wrapped in a control problem, and the two problems must be tackled simultaneously for good data efficiency. This is because the policy of the agent affects the data it will collect, which in turn affects the policy, and so on. This is in contrast to supervised learning, where the performance of a classifier (for instance) does not influence the data it will later observe. Failure to properly consider the statistical aspect of the RL problem will result in agents that require exponential amounts of experience for good performance. On the other hand, correctly considering the statistical inference problem and the control problem together has the potential to dramatically reduce the compute requirements to solve problems and potentially unlock new domains and capabilities far outside of the range of current agents. In this talk I will introduce these concepts and discuss how Bayesian techniques can provide principled solutions to the problem. Speaker Biography Brendan O'Donoghue earned his PhD in 2013 from Stanford working with Stephen Boyd on optimization and control theory. Since then he has worked at DeepMind as a research scientist working on deep reinforcement learning, optimization, and (more recently) large language models. Please check the seminar page https://engineering.oregonstate.edu/EECS/research/AI-seminars<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> for future talks. Prasad Tadepalli

1 0

information theory class
by Nguyen, Thinh 05 Jan '24

05 Jan '24

Hello everyone, I just want to bring to your attention that I will be teaching information theory class (ECE 566) next term. If you are interested in learning the theories and some foundational connections among various disciplines (compression, compression, machine learning, statistics, ...) then you might want to consider taking ECE 566. I am attaching the ECE 566 syllabus. If enough number of student register for ECE 566, I will offer the second course (special topics) in information theory series in the Spring term. Best, Dr. Nguyen

1 1

SP virtual Seminar: Generative AI for Coronary Imaging (11 am, Dec 19)
by Fu, Xiao 11 Dec '23

11 Dec '23

Hi all, Please mark your calendar for the upcoming seminar before the holidays. The zoom link will be sent later today. Best, Xiao Time: 11 am Dec 19, 2023 Title: Generative AI for coronary imaging Abstract: Coronary artery disease (CAD) has been influencing a market over $2.8 billion with roughly 1,000,000 treatment procedures performed annually. Existing guidance of CAD treatment suffers from the limited spatial resolution and lacks real-time detailed pathological identification. Even the finest coronary imaging system - intravascular optical coherence tomography (IVOCT), is not able to resolve the structure at sub-cellular level and vessel characterization at sub-cellular level can only be obtained after patient’s death through histopathology. We will discuss some examples in our research team of developing a data-driven approach using generative artificial intelligence (AI) to use off-line data and training process, without any hardware modifications, to generatively produce new information that could not be acquired previously or can only be obtained ex vivo. We will also cover some latest progress in developing robust algorithms to segment medical images, including OCT images. Bio: Dr. Yu Gan is currently an Assistant Professor in the Biomedical Engineering Department at Stevens Institute of Technology. He received his Ph. D. degree in Electrical Engineering from Columbia University in 2017. Prior to Stevens, He was an Assistant Professor in the Electrical and Computer Engineering at the University of Alabama from 2018 to 2021. Dr. Gan’s research lies in machine learning and deep learning techniques that facilitate biomedical imaging and dig out hidden information from medical/biomedical images. His research has been supported by National health institute (NIH), NJ Health Foundations, National science foundation (NSF), US Department of Agriculture (USDA), Burrough Wellcome foundation, and other agency (companies). He is the recipient of 2023 NSF CAREER award. ==== Xiao Fu, Associate Professor School of Electrical Engineering and Computer Science Oregon State University Corvallis, OR 97331 http://eecs.oregonstate.edu/~fuxia<https://nam04.safelinks.protection.outlook.com/?url=http%3A%2F%2Fpeople.ore…>

1 1

AI Seminar : Dec 08
by Mangannavar, Rajesh Devaraddi 07 Dec '23

07 Dec '23

Dear all, Our next AI seminar is scheduled to be on December 8th, 1-2 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> AI-related Research in Education and Counseling Cass Dykeman Professor of Counseling College of Education Oregon State University Abstract: In various professions, Artificial Intelligence has significantly influenced both research methodologies and subjects. The fields of Education and Counseling are no exceptions. This presentation will delve into how researchers at OSU’s College of Education utilize constituent elements of AI, including the analysis of: (a) collocations, (b) n-grams/skip-grams, (c) authorial stance, (d) keyness, (e) sentiment, and (f) topic models. Furthermore, I will highlight AI research tools beneficial for graduate students and faculty, such as Noteable.io, Elicit.com, Claude.ai, and Orange Data Mining. We will also address the ethical considerations surrounding the use of AI tools in crafting IMRaD manuscripts intended for submission to peer-reviewed journals. Speaker Bio: Cass Dykeman is a professor of counseling in OSU’s College of Education who received his Ph.D. in counselor education from the University of Virginia and a master's in educational psychology from the University of Washington. Prior to working in higher education, Dykeman served as an elementary school counselor in Seattle. At OSU since 1998, he holds the record for the number of dissertations completed (60), having passed James D. White of Chemistry in 2022. The last 25 of these dissertations have employed AI-related research methods. A former Army officer, Dykeman’s own research focuses on AI-augmented counterintelligence. Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 0

AI Seminar : Dec 1
by Mangannavar, Rajesh Devaraddi 29 Nov '23

29 Nov '23

Dear all, Our next AI seminar is scheduled to be on December 1stth, 1-2 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Best Practices for Robot Death Heather Knight Assistant Professor Computer Science Oregon State University Abstract: Have you ever been sad after losing a pet, lost all data on your phone, or spilled coffee on your laptop? Imagine how a robot owner might feel after losing their companion robot, or seeing its memory and capability erased. We often replace phones after mere years of use, robot software goes out of date much faster than a dog or cat's lifespan, yet there are few social considerations for the grief a human might feel in losing software support for a robot system, or over robot hardwares for which no new parts are constructed. Yet, just like living systems, impermanence in technological systems is a given. This talk considers ways in which perspectives from Social Robotics — a field that leverages anthropomorphism, i.e., the ways in which humans automatically imbue humanlike characteristics to non-human things — can inform future social considerations and best practices for Robot Death. For example, social structures and legal mechanisms that take human-robot bonding into account by compassionately delivering news, and/or validating naturally existing sadness post-machine decline. Drawing from my personal experience having a Nao robot stolen out of my car in Corvallis, OR, from analyses of robot funerals after Starship Robots are hit by cars or trains, and international examples of ceremonies conducted for lost machines, this talk acknowledges that digital lives can also experience existential 1's and 0's. What can Social Robotics as a field teach us about compassionately handling robot and machine death? Moreover, how might robots and machines help us to compassionately navigate impermanence and suffering in our own lives? We finish with a discussion of opportunities for robot-facilitated mindfulness practices, including "maranasati," which means mindfulness of death and has surprising benefits to one's mental health. Speaker Bio: Dr. Heather Knight is an Assistant Professor of Computer Science at Oregon State University, and directs the CHARISMA Robotics Lab, advising students in Robotics, EECS, AI, and MIME. Her students investigate multi-robot expressive motion, human-robot interactive communication, interfaces for human-in-the-loop robot control, and frequently use entertainment methods to bootstrap the development of everyday social robots. Past projects include Cyberflora, a robot flower garden installation at the Smithsonian/Cooper-Hewitt Design Museum, Silicon Comedy, an online learning system for interactive robot comedy featured on TED.com, and This Too Shall Pass, a music video featuring a two-floor Rube Goldberg Machine that won a British Video Music Award. Her academic background included a postdoc at Stanford University exploring minimal robots and autonomous car interfaces, a Robotics Ph.D. at Carnegie Mellon University on Expressive Motion for Low Degree-of-Freedom Robots, and M.S. and B.S. in Electrical Engineering & Computer Science at Massachusetts Institute of Technology, where she developed a sensate skin for a robot teddy bear at the MIT Media Lab. Outside of academia, she has developed robotics and instrumentation at NASA’s Jet Propulsion Laboratory, and sensor and field applications as an early employee of Aldebaran Robotics, the developers of the Nao and Pepper robots (since acquired by Softbank). Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 0

AI Seminar : Nov 17
by Mangannavar, Rajesh Devaraddi 15 Nov '23

15 Nov '23

Dear all, Our next AI seminar is scheduled to be on Nov 17th, 1-2 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> New Frontiers In Adaptive Experimental Design For Multi-objective Optimization Syrine Belakaria Data Science Postdoctoral Fellow Computer Science Stanford University Abstract: Many design optimization problems in science, engineering, and industrial domains are instantiations of the following general problem: adaptive optimization of complex design spaces guided by expensive experiments where the expense is measured in terms of resources consumed by the experiments. This talk focuses on the problem of multi-objective optimization (MOO) using expensive black-box function evaluations (also referred to as experiments) where the goal is to approximate the optimal Pareto set of solutions by minimizing the total resource cost of experiments. For example, in drug design optimization, we need to find designs that trade off effectiveness, safety, and cost using expensive experiments. The key challenge is to select the sequence of experiments to uncover high-quality solutions by minimizing the total resource cost. In this talk, I will describe a general framework for solving MOO problems based on the output space entropy (OSE) search principle: select the experiment that maximizes the information gained per unit resource cost about the optimal Pareto front. I will also explain how to instantiate the principle of OSE search to derive efficient algorithms for the following four MOO problem settings: 1) The most basic single-fidelity setting where experiments are expensive and accurate; 2) Handling black-box constraints that cannot be evaluated without performing experiments; 3) The multi-fidelity setting where candidate experiments can vary in the amount of resources consumed and their evaluation accuracy. I will present experimental results on real-world engineering and science applications to demonstrate the effectiveness of OSE framework in terms of accuracy of MOO solutions and computational efficiency. Speaker Bio: Syrine Belakaria is a Data Science Postdoctoral fellow in the Computer Science department at Stanford University working with Professor Stefano Ermon and Professor Barbara Engelhardt. She obtained her PhD in Computer Science from Washington State University where she was advised by Professor Jana Doppa; MS in Electrical Engineering from the University of Idaho; and Engineering degree in Information Technology from the Higher School of Communication of Tunis, Tunisia. She won IBM PhD Fellowship (2021-2023) and was a Finalist for Microsoft Research Fellowship (2021), was selected for the MIT Rising Stars in EECS (2021), won the WSU Harriet Rigas Outstanding Woman in Doctoral Studies Award (2023), won Outstanding TA award in CS (2019), and two Outstanding Reviewer Awards from the ICML conference. She spent time as a research intern at Microsoft Research and Meta Research. Her general research interests are in the broad area of AI for science and engineering with a current focus on adaptive experiment design to solve real-world problems including hardware design, materials design, electric transportation systems, and Auto ML. Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 0

AI Seminar : Nov 10
by Mangannavar, Rajesh Devaraddi 08 Nov '23

08 Nov '23

Dear all, Our next AI seminar on "New Frontiers In Adaptive Experimental Design For Multi-objective Optimization" by Syrine Belakaria is scheduled to be on Nov 10th (Friday), 1-2 PM. It will be followed by a 30-minute Q&A session with the graduate students. Location: KEC 1001 Zoom link: https://oregonstate.zoom.us/j/98684050301?pwd=ZzhianQxUFBPUmdYVWJKOFhaVURCQ…<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Foregonsta…> Learning Interpretable Models on Complex Medical Data Syrine Belakaria Data Science Postdoctoral Fellow Computer Science Stanford University Abstract: Many design optimization problems in science, engineering, and industrial domains are instantiations of the following general problem: adaptive optimization of complex design spaces guided by expensive experiments where the expense is measured in terms of resources consumed by the experiments. This talk focuses on the problem of multi-objective optimization (MOO) using expensive black-box function evaluations (also referred to as experiments) where the goal is to approximate the optimal Pareto set of solutions by minimizing the total resource cost of experiments. For example, in drug design optimization, we need to find designs that trade off effectiveness, safety, and cost using expensive experiments. The key challenge is to select the sequence of experiments to uncover high-quality solutions by minimizing the total resource cost. In this talk, I will describe a general framework for solving MOO problems based on the output space entropy (OSE) search principle: select the experiment that maximizes the information gained per unit resource cost about the optimal Pareto front. I will also explain how to instantiate the principle of OSE search to derive efficient algorithms for the following four MOO problem settings: 1) The most basic single-fidelity setting where experiments are expensive and accurate; 2) Handling black-box constraints that cannot be evaluated without performing experiments; 3) The multi-fidelity setting where candidate experiments can vary in the amount of resources consumed and their evaluation accuracy. I will present experimental results on real-world engineering and science applications to demonstrate the effectiveness of OSE framework in terms of accuracy of MOO solutions and computational efficiency. Speaker Bio: Syrine Belakaria is a Data Science Postdoctoral fellow in the Computer Science department at Stanford University working with Professor Stefano Ermon and Professor Barbara Engelhardt. She obtained her PhD in Computer Science from Washington State University where she was advised by Professor Jana Doppa; MS in Electrical Engineering from the University of Idaho; and Engineering degree in Information Technology from the Higher School of Communication of Tunis, Tunisia. She won IBM PhD Fellowship (2021-2023) and was a Finalist for Microsoft Research Fellowship (2021), was selected for the MIT Rising Stars in EECS (2021), won the WSU Harriet Rigas Outstanding Woman in Doctoral Studies Award (2023), won Outstanding TA award in CS (2019), and two Outstanding Reviewer Awards from the ICML conference. She spent time as a research intern at Microsoft Research and Meta Research. Her general research interests are in the broad area of AI for science and engineering with a current focus on adaptive experiment design to solve real-world problems including hardware design, materials design, electric transportation systems, and Auto ML. Please watch this space for future AI Seminars : https://engineering.oregonstate.edu/EECS/research/AI<https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fengineeri…> Rajesh Mangannavar, Graduate Student Oregon State University ---- AI Seminar Important Reminders: -> For graduate students in the AI program, attendance is strongly encouraged

1 1