Select a calendar:
Filter April Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for April
-
CS Colloquium: Paul Schmitt (USC ISI) - Networked Systems for a Modern, Private Internet
Mon, Apr 04, 2022 @ 10:00 AM - 11:00 AM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Paul Schmitt, USC ISI
Talk Title: Networked Systems for a Modern, Private Internet
Abstract: Users expect that the networks and protocols they use protect their privacy. Unfortunately, many ubiquitous legacy systems have significant privacy flaws. Network operators face a different challenge: widespread adoption of encryption, while a clear benefit to users, reduces operator visibility into traffic flowing through their networks. In this talk, I discuss networked systems to enhance user privacy and systems and techniques for privacy-preserving network traffic analysis. I describe my research that leverages key architectural points of decoupling to enhance privacy in the global DNS ecosystem and in mobile networks. I then discuss systems I have built for privacy-preserving network analysis for use by network operators to gain insight into network usage and performance, all without breaking encryption.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Paul Schmitt is a research computer scientist at ISI. His research areas include networked systems, privacy, network traffic inference and analysis, and scalable Internet measurement. His work takes a dirty-slate approach to networked systems research, allowing for compatibility and immediate deployability in current environments. He previously received his PhD from UC Santa Barbara in 2017 and was an associate research scholar at Princeton University.
Host: John Heidemann
Audiences: Everyone Is Invited
Contact: Cherie Carter
-
CAIS Seminar: David Eddie (Massachusetts General Hospital) - Towards a biosensor-driven, just-in-time relapse prevention tool for substance use disorder: Identifying neurocardiac biomarkers of stress and relapse risk
Mon, Apr 04, 2022 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: David Eddie, Massachusetts General Hospital
Talk Title: Towards a biosensor-driven, just-in-time relapse prevention tool for substance use disorder: Identifying neurocardiac biomarkers of stress and relapse risk
Series: USC Center for Artificial Intelligence in Society (CAIS) Seminar Series
Abstract: Substance use disorders carry tremendous personal and societal costs, and despite best patient and clinical efforts, relapse is common. Much research has sought to identify psychosocial risk factors for addiction relapse, but much less attention has been paid to how psychophysiological impairment may confer risk. In this talk, I will highlight how stress and central autonomic network dysregulation reflected by reduced heart rate variability (HRV) may heighten risk for individuals in early alcohol use disorder (AUD) recovery, showing that HRV can be used to predict subsequent alcohol use. I will also show preliminary findings from a study that aims to use smartwatches and machine learning to identify stress states, with the goal of developing a just-in-time relapse prevention tool for individuals in early recovery from substance use disorder.
Register in advance for this webinar at:
https://usc.zoom.us/webinar/register/WN_Hcft0t87RQqrca66W5c8ug
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: David Eddie, Ph.D. is a research scientist at Massachusetts General Hospital's Recovery Research Institute and Center for Addiction Medicine, a clinical psychologist in Massachusetts General Hospital's Department of Psychiatry, and an assistant professor at Harvard Medical School. His current projects include an NIAAA supported study developing a biosensor driven just-in-time intervention for substance use disorders, and a NIDA supported project assessing the efficacy of a novel mutual-help addiction recovery program based on physical activity.
Host: USC Center for Artificial Intelligence in Society (CAIS)
Webcast: https://usc.zoom.us/webinar/register/WN_Hcft0t87RQqrca66W5c8ugLocation: Online - Zoom Webinar
WebCast Link: https://usc.zoom.us/webinar/register/WN_Hcft0t87RQqrca66W5c8ug
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Suguman Bansal (University of Pennsylvania) - Specification-Guided Policy Synthesis
Tue, Apr 05, 2022 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Suguman Bansal , University of Pennsylvania
Talk Title: Specification-Guided Policy Synthesis
Series: CS Colloquium
Abstract: Policy synthesis or algorithms to design policies for computational systems is one of the fundamental problems in computer science. Standing on the shoulders of simplified yet concise task-specification using high-level logical specification languages, this talk will cover synthesis algorithms using two contrasting approaches. First, the classical logic-based approach of reactive synthesis; Second, the modern learning-based approach of reinforcement learning. This talk will cover our scalable and efficient state-of-the-art algorithms for synthesis from high-level specifications using both these approaches, and investigate whether formal guarantees are possible. We will conclude with a forward-looking view of these contributions to trustworthy AI.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Suguman Bansal is an NSF/CRA Computing Innovation Postdoctoral Fellow at the University of Pennsylvania, mentored by Prof. Rajeev Alur. Her primary area of research is Formal Methods and Programming Langauge, and her secondary area of research is Artificial Intelligence.
https://suguman.github.io/
Host: Mukund Raghothaman
Location: Olin Hall of Engineering (OHE) - 132
Audiences: By invitation only.
Contact: Assistant to CS chair
-
CS Colloquium: Daniel Grier (University of Waterloo) - The Complexity of Near-Term Quantum Computers
Tue, Apr 05, 2022 @ 02:00 PM - 03:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Daniel Grier, University of Waterloo
Talk Title: The Complexity of Near-Term Quantum Computers
Series: CS Colloquium
Abstract: Quantum computing is at an exciting moment in its history, with some high-profile experimental successes in building programmable quantum devices. That said, these quantum devices (at least in the near term) will be restricted in several ways, raising questions about their power relative to classical computers. In this talk, I will present three results which give us a better understanding of these near-term quantum devices, revealing key features which make them superior to their classical counterparts.
First, I will show that constant-depth quantum circuits can solve problems that cannot be solved by any constant-depth classical circuit consisting of AND, OR, NOT, and PARITY gates---giving the largest-known unconditional separation between natural classes of quantum and classical circuits. Second, I will show that these quantum circuits can nevertheless be simulated quickly by classical algorithms that have no depth restriction, emphasizing the role that depth plays in provable quantum advantage. Finally, I will address some of the experimental challenges in implementing linear optical quantum computers, and I will prove that they outperform classical computers using standard conjectures but in more practical experimental regimes.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Daniel is a postdoctoral researcher at the Institute for Quantum Computing at the University of Waterloo. He received his PhD in Computer Science at MIT, where he was advised by Scott Aaronson and was supported by an NSF Graduate Research Fellowship. His research lies at the intersection of complexity theory and quantum computing, with a particular focus on the power of near-term quantum computing devices.
Host: Ramesh Govindan
Location: online only
Audiences: By invitation only.
Contact: Assistant to CS chair
-
CS Colloquium: Geoff Pleiss (Columbia University) - Bridging the Gap Between Deep Learning and Probabilistic Modeling
Thu, Apr 07, 2022 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Geoff Pleiss , Columbia University
Talk Title: Bridging the Gap Between Deep Learning and Probabilistic Modeling
Series: CS Colloquium
Abstract: Deep learning excels with large-scale unstructured data - common across many modern application domains - while probabilistic modeling offers the ability to encode prior knowledge and quantify uncertainty - necessary for safety-critical applications and downstream decision-making tasks. I will discuss examples from my research that bridge the gap between these two learning paradigms. The first half will show that insights from deep learning can improve the practicality of probabilistic models. I will discuss work that scales Gaussian process regression, a common probabilistic model, to datasets two orders of magnitude larger than previously reported. The second half will show that probabilistic methods can improve our understanding of deep learning. I will demonstrate that Gaussian process theory uncovers new insights about the effects of width and depth in neural networks. I will conclude with ongoing efforts to quantify neural network uncertainty, develop new inductive biases, and other work at the intersection of deep learning and probabilistic modeling.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Geoff Pleiss is a postdoctoral researcher at Columbia University, hosted by John Cunningham, with affiliations in the Department of Statistics and the Zuckerman Institute. He obtained his Ph.D. in Computer Science from Cornell University, advised by Kilian Weinberger, and his B.Sc. from Olin College of Engineering. His research interests are broadly situated in machine learning, including neural networks, Gaussian processes, uncertainty quantification, and scalability. Geoff is also the co-founder and maintainer of the GPyTorch software framework.
Host: Robin Jia
Location: Olin Hall of Engineering (OHE) - 132
Audiences: By invitation only.
Contact: Assistant to CS chair
-
CS Colloquium: Hussein Sibai (UC Berkeley) - Towards Physics-aware Trustworthy Autonomy
Thu, Apr 07, 2022 @ 02:00 PM - 03:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Hussein Sibai , UC Berkeley
Talk Title: Towards Physics-aware Trustworthy Autonomy
Series: CS Colloquium
Abstract: Designing trustworthy autonomous systems is a looming challenge in several domains. Symbolic reasoning and verification can complement purely data-driven approaches by exploiting knowledge of structure and code, providing rigorous safety assurances, explaining why designs work, and helping find edge-cases quickly. In this talk, I will discuss recent results that use knowledge about physical laws, such as symmetries, to boost the scalability of formal verification of autonomous systems. The boosting benefits both data-driven and model-based analysis. My tool SceneChecker embodies these algorithms and data structures that use knowledge of symmetries to save verification algorithms from repeating expensive reachability computations. It implements a counterexample-guided abstraction-refinement (CEGAR) verification algorithm that compresses models by combining symmetric states. SceneChecker has been successful in verifying complex scenarios involving ground and aerial vehicles. In the second half, I will present results developed using notions from topological entropy to relate knowledge of physical laws governing a system with data requirements in solving estimation and verification problems. These results can give physics-aware lower-bounds that can guide future autonomy design processes.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Hussein Sibai is a Postdoctoral Scholar at UC Berkeley, advised by Murat Arcak and Sanjit Seshia. He obtained his Ph.D. in Electrical and Computer Engineering from the University of Illinois Urbana-Champaign (UIUC) in December 2021, advised by Sayan Mitra. He received his bachelor's degree in Computer and Communication Engineering from the American University of Beirut and a master's degree in Electrical and Computer Engineering from UIUC. His research interests are in formal methods, control theory, and machine learning. Hussein has won the best poster award in HSCC 2018 and best paper nominations at HSCC 2017 and ATVA 2019. His work has been recognized by the Rambus fellowship, the Ernest A. Reid fellowship, the MAVIS Future Faculty fellowship, and the ACM SIGBED gold medal for the graduate category in the student research competition in CPS Week 21.
Host: Jyo Deshmukh
Location: Ronald Tutor Hall of Engineering (RTH) - 526
Audiences: By invitation only.
Contact: Assistant to CS chair
-
CS Colloquium: Priya Donti (Carnegie Mellon University) - Optimization-in-the-loop AI for energy and climate
Fri, Apr 08, 2022 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Priya Donti , Carnegie Mellon University
Talk Title: Optimization-in-the-loop AI for energy and climate
Series: CS Colloquium
Abstract: Addressing climate change will require concerted action across society, including the development of innovative technologies. While methods from artificial intelligence (AI) and machine learning (ML) have the potential to play an important role, these methods often struggle to contend with the physics, hard constraints, and complex decision-making processes that are inherent to many climate and energy problems. To address these limitations, I present the framework of "optimization-in-the-loop AI," and show how it can enable the design of AI models that explicitly capture relevant constraints and decision-making processes. For instance, this framework can be used to design learning-based controllers that provably enforce the stability criteria or operational constraints associated with the systems in which they operate. It can also enable the design of task-based learning procedures that are cognizant of the downstream decision-making processes for which a model's outputs will be used. By significantly improving performance and preventing critical failures, such techniques can unlock the potential of AI and ML for operating low-carbon power grids, improving energy efficiency in buildings, and addressing other high-impact problems of relevance to climate action.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Priya Donti is a Ph.D. Candidate in Computer Science and Public Policy at Carnegie Mellon University. Her research explores methods to incorporate physics and hard constraints into deep learning models, in order to enable their use for forecasting, optimization, and control in high-renewables power grids. She is also a co-founder and chair of Climate Change AI, an initiative to catalyze impactful work in climate change and machine learning. Priya is a recipient of the MIT Technology Review's 2021 "35 Innovators Under 35" award, the Siebel Scholarship, the U.S. Department of Energy Computational Science Graduate Fellowship, and best paper awards at ICML (honorable mention), ACM e-Energy (runner-up), PECI, the Duke Energy Data Analytics Symposium, and the NeurIPS workshop on AI for Social Good.
Host: Bistra Dilkina
Location: Ronald Tutor Hall of Engineering (RTH) - 115
Audiences: By invitation only.
Contact: Assistant to CS chair
-
CAIS Seminar: Rayid Ghani (Carnegie Mellon University) - Practical Lessons and Challenges in Building Fair and Equitable AI/ML Systems
Mon, Apr 11, 2022 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Rayid Ghani, Carnegie Mellon University
Talk Title: Practical Lessons and Challenges in Building Fair and Equitable AI/ML Systems
Series: USC Center for Artificial Intelligence in Society (CAIS) Seminar Series
Abstract: As organizations become more aware of the need to build ML/AI systems that result in fair and equitable outcomes, they have started to struggle with operationalizing that need. In this talk, I'll discuss lessons learned over the past few years working with various government agencies and non-profits across health, criminal justice, social services, education, and economic & workforce development on how those organizations view this challenge, how they're attempting to design ML/AI systems, and what gaps exist in the work that Fair ML researchers have been producing. I'll also discuss some examples of methods and tools that were useful in those collaborations and resulted in more equitable impact through the use of ML.
Register in advance for this webinar at:
https://usc.zoom.us/webinar/register/WN_zr43DpG2SKaIj-rspOahZA
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Rayid Ghani is a Professor in Machine Learning and Public Policy at Carnegie Mellon University focused on developing and using AI/Machine Learning/Data Science to help tackle large public policy and societal challenges in a fair and equitable manner. Among other areas, Rayid works with governments and non-profits in policy areas such as health, criminal justice, education, public safety, economic development, and urban infrastructure. Before joining Carnegie Mellon University, Rayid was the Founding Director of the Center for Data Science & Public Policy, Research Associate Professor in Computer Science, and a Senior Fellow at the Harris School of Public Policy at the University of Chicago. Previously, Rayid was the Chief Scientist of the Obama 2012 Election Campaign.
Host: USC Center for Artificial Intelligence in Society (CAIS)
Webcast: https://usc.zoom.us/webinar/register/WN_zr43DpG2SKaIj-rspOahZALocation: Online - Zoom Webinar
WebCast Link: https://usc.zoom.us/webinar/register/WN_zr43DpG2SKaIj-rspOahZA
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
Remarkable Trajectory Lecture: Paul S. Rosenbloom (USC) - From Designing Minds to Mapping Disciplines
Tue, Apr 12, 2022 @ 02:00 PM - 03:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Paul S. Rosenbloom, University of Southern California
Talk Title: From Designing Minds to Mapping Disciplines
Series: Remarkable Trajectory Lecture Series
Abstract: Designing minds involves understanding the fixed mechanisms that combine to yield a mind as a basis for building both integrated models of human cognition and general AI systems. My trajectory here began in the mid-to-late 1970s with rule-based systems, and evolved through a sequence of more elaborate cognitive architectures -“ Xaps, Soar, and Sigma. It has also included recent efforts to understand minds more abstractly, in terms of a Common Model of Cognition and dichotomic maps of architectural mechanisms. Mapping disciplines involves understanding their essences and systematically structuring their compositions. My trajectory here began with a relational map of computing as a great scientific domain and continued with recent work on dichotomic maps of the technologies underlying AI and cognitive science. Following a dab of personal background, I will overview these two trajectories, and then wrap up with a bit of speculation on their affinity and a sampling of maxims extracted from my career as a whole.
Register in advance for this online event at:
https://usc.zoom.us/webinar/register/WN__4hJussyRBus_HIFLcgigQ
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Paul S. Rosenbloom recently retired as a Professor of Computer Science in the Viterbi School of Engineering at the University of Southern California and Director for Cognitive Architecture Research at the Institute for Creative Technologies. He also was a member of USC's Information Sciences Institute for two decades, ending as its deputy director, and earlier was faculty at Carnegie Mellon University and Stanford University (with a joint appointment in Computer Science and Psychology). His research has focused on cognitive architectures (models of the fixed structures and processes that together yield a mind), the Common Model of Cognition (a partial consensus about the structure of a human-like mind), dichotomic maps (structuring the space of technologies underlying AI and cognitive science), and the relational model of computing as a great scientific domain (akin to the physical, life and social sciences). He is a Fellow of the Association for the Advancement of Artificial Intelligence (AAAI), the American Association for the Advancement of Science (AAAS), and the Cognitive Science Society; and with John E. Laird was awarded the Herbert A. Simon Prize for Advances in Cognitive Systems. He has served as Councilor and Conference Chair for AAAI; Chair of ACM SIGART (now SIGAI); Chair of the Viterbi Engineering Faculty Council; and President of the USC Faculty.
Host: USC Viterbi School of Engineering Department of Computer Science
Webcast: https://usc.zoom.us/webinar/register/WN__4hJussyRBus_HIFLcgigQLocation: Online - Zoom Webinar
WebCast Link: https://usc.zoom.us/webinar/register/WN__4hJussyRBus_HIFLcgigQ
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Bradley Hayes (University of Colorado Boulder) - Human-robot teaming is a lot less dangerous with communication: Improving Human-Robot Teaming Performance in Partially Observable Environments with Augmented Reality
Wed, Apr 13, 2022 @ 04:30 PM - 05:50 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Bradley Hayes, University of Colorado Boulder
Talk Title: Human-robot teaming is a lot less dangerous with communication: Improving Human-Robot Teaming Performance in Partially Observable Environments with Augmented Reality
Series: Computer Science Colloquium
Abstract: Clear and frequent communication is a foundational aspect of collaboration. Effective communication not only enables and sustains the shared situational awareness necessary for adaptation and coordination during human-robot teaming, but is often a requirement given the opaque nature of decision-making in autonomous systems. In this talk I will share some of our recent work using augmented reality as a mode of visual communication to improve both human and robot safety and capability when working together, introducing insights into human behavior and compliance in safety-critical situations as well as novel algorithms for autonomous communication and collaboration in partially observable environments. The talk will conclude with a presentation of our ongoing work at the intersection of fast constrained motion planning for sequential manifold planning problems and augmented reality-assisted learning from demonstration.
Prof. Bradley Hayes will give his talk in person at GFS 106 and we will also host the talk over Zoom.
Register in advance for this webinar at:
https://usc.zoom.us/webinar/register/WN_HgvCIbb7TDS6aOU1ksSI0A
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Bradley Hayes is an Assistant Professor of Computer Science at the University of Colorado Boulder, where he runs the Collaborative AI and Robotics (CAIRO) Lab and serves as co-director of the university's Autonomous Systems Interdisciplinary Research Theme. Brad's research develops techniques to create and validate autonomous systems that learn from, teach, and collaborate with humans to improve efficiency, safety, and capability at scale. His work primarily leverages novel approaches at the intersection of human-robot interaction and explainable artificial intelligence, providing autonomous systems with the ability to generalize skills with limited risk, to act safely and productively around humans, and to make human-autonomy teams more powerful than the sums of their parts. His continual efforts to systematically put humans and autonomous systems into often entertaining and occasionally productive situations has been featured by TEDx, Popular Science, Wired, and MIT Technology review, and has been recognized with best paper nominations from HRI, AAMAS, and RO-MAN. Brad also serves as CTO at Circadence, building high-fidelity simulation, test, and evaluation environments for cyber-physical systems at nation-state scale.
Host: Stefanos Nikolaidis
Webcast: https://usc.zoom.us/webinar/register/WN_HgvCIbb7TDS6aOU1ksSI0ALocation: Grace Ford Salvatori Hall Of Letters, Arts & Sciences (GFS) - 106
WebCast Link: https://usc.zoom.us/webinar/register/WN_HgvCIbb7TDS6aOU1ksSI0A
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Joydeep Biswas (University of Texas at Austin) - Deploying Autonomous Service Mobile Robots, And Keeping Them Autonomous
Thu, Apr 14, 2022 @ 04:10 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Joydeep Biswas, University of Texas at Austin
Talk Title: Deploying Autonomous Service Mobile Robots, And Keeping Them Autonomous
Series: Computer Science Colloquium
Abstract: *New start time: 4:10 PM PT*
Why is it so hard to deploy autonomous service mobile robots in unstructured human environments, and to keep them autonomous? In this talk, I will explain three key challenges, and our recent research in overcoming them: 1) ensuring robustness to environmental changes; 2) anticipating and overcoming failures; and 3) efficiently adapting to user needs.
To remain robust to environmental changes, we build probabilistic perception models to explicitly reason about object permanence and distributions of semantically meaningful movable objects. By anticipating and accounting for changes in the environment, we are able to robustly deploy robots in challenging frequently changing environments.
To anticipate and overcome failures, we introduce introspective perception to learn to predict and overcome perception errors. Introspective perception allows a robot to autonomously learn to identify causes of perception failure, how to avoid them, and how to learn context-aware noise models to overcome such failures.
To adapt and correct behaviors of robots based on user preferences, or to handle unforeseen circumstances, we leverage representation learning and program synthesis. We introduce visual representation learning for preference-aware planning to identify and reason about novel terrain types from unlabelled human demonstrations. We further introduce physics-informed program synthesis to synthesize and repair programmatic action selection policies (ASPs) in a human-interpretable domain-specific language with several orders of magnitude fewer demonstrations than necessary for neural network ASPs of comparable performance.
The combination of these research advances allows us to deploy a varied fleet of wheeled and legged autonomous mobile robots on the campus scale at UT Austin, performing tasks that require robust mobility both indoors and outdoors.
***Dr. Joydeep Biswas will give the talk in person at SGM 124 and we will also host the talk over Zoom.***
Register in advance for this webinar at:
https://usc.zoom.us/webinar/register/WN_lWf_mXH3Qr2qtbHg1kbOYQ
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Joydeep Biswas is an assistant professor in the department of computer science at the University of Texas at Austin. He earned his B.Tech in Engineering Physics from the Indian Institute of Technology Bombay in 2008, and M.S. and PhD in Robotics from Carnegie Mellon University in 2010 and 2014 respectively. From 2015 to 2019, he was assistant professor in the College of Information and Computer Sciences at the University of Massachusetts Amherst. His research spans perception and planning for long-term autonomy, with the ultimate goal of having service mobile robots deployed in human environments for years at a time, without the need for expert corrections or supervision. Prof. Biswas received the NSF CAREER award in 2021, an Amazon Research Award in 2018, and a JP Morgan Faculty Research Award in 2018.
Host: Stefanos Nikolaidis
Webcast: https://usc.zoom.us/webinar/register/WN_lWf_mXH3Qr2qtbHg1kbOYQLocation: Seeley G. Mudd Building (SGM) - 124
WebCast Link: https://usc.zoom.us/webinar/register/WN_lWf_mXH3Qr2qtbHg1kbOYQ
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Mohamed Hussein (USC ISI) - Securing Machine Vision Models
Fri, Apr 15, 2022 @ 10:00 AM - 11:00 AM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Mohamed Hussein, USC ISI
Talk Title: Securing Machine Vision Models
Abstract: Machine vision has evolved dramatically over the past decade, thanks to the deep learning revolution. Despite their remarkable performance, often surpassing humans, machine vision models are vulnerable to different types of attacks. This talk will focus on two types of attacks as well as methods to secure machine vision models against them. The first is presentation (or more commonly known as spoofing) attacks on biometric authentication systems, in which the attacker presents a fake physical instrument to the system, such as a printed face image, either to conceal their true identity or impersonate a different identity. I will show that combining the power of deep learning with multi-spectral sensing can effectively address this problem by distinguishing spoofing instruments from bona fide presentations. For the challenging makeup attack, I will show that using multi-spectral data, we can construct an image of a person without the applied makeup, and hence reveal their true identity. The second type of attack is adversarial attacks. In this type of attack, imperceptible perturbations can be applied to the input of a machine vision model to alter the model's prediction. I will present a new non-linear activation function, named Difference of Mirrored Exponential terms (DOME), which has the property of inducing compactness to the embedding space of a deep learning model. We found that combining the usage of DOME with adversarial training can boost the robustness against state of the art adversarial attacks. I will conclude by discussing my perspective on the challenges ahead regarding the security of machine vision models.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Dr. Mohamed E. Hussein is a Computer Scientist and a Research Lead at USC ISI. Dr. Hussein obtained his Ph.D. degree in Computer Science from the University of Maryland at College Park, MD, USA in 2009. Then, he spent close to two years as an Adjunct Member Research Staff at Mitsubishi Electric Research Labs, Cambridge, MA, before moving to Alexandria University, Egypt, as a faculty member. Prior to joining ISI in 2017, he spent three years at Egypt-Japan University of Science and Technology (E-JUST), Alexandria, Egypt. During his time as a faculty member in Egypt, Dr. Hussein was the PI/Co-PI on multiple industry and government funded research projects on Sign Language Recognition and Crowd Scene Analysis. He is currently a Co-PI for ISI's projects under IARPA's Odin and BRIAR programs and DARPA's GARD program.
Host: CS Department
Webcast: https://usc.zoom.us/j/98761669161Location: Ronald Tutor Hall of Engineering (RTH) - 105
WebCast Link: https://usc.zoom.us/j/98761669161
Audiences: Everyone Is Invited
Contact: Cherie Carter
-
CS Colloquium: Beidi Chen (Stanford University) - Randomized Algorithms for Efficient Machine Learning Systems
Fri, Apr 15, 2022 @ 02:00 PM - 03:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Beidi Chen, Stanford University
Talk Title: Randomized Algorithms for Efficient Machine Learning Systems
Series: CS Colloquium
Abstract: Machine learning (ML) has demonstrated great promise in scientific discovery, healthcare, and education, especially with the rise of large neural networks. However, large models trained on complex and rapidly growing data consume enormous computational resources. In this talk, I will describe my work on exploiting model sparsity with randomized algorithms to accelerate large ML systems on current hardware with no drop in accuracy.
I will start by describing SLIDE, an open-source system for efficient sparse neural network training on CPUs that has been deployed by major technology companies and academic labs. SLIDE blends Locality Sensitive Hashing with multi-core parallelism and workload optimization to drastically reduce computations. SLIDE trains industry-scale recommendation models on a 44 core CPU 3.5x faster than TensorFlow on V100 GPU with no drop in accuracy.
Next, I will present Pixelated Butterfly, a simple yet efficient sparse training framework on GPUs. It uses a simple static block-sparse pattern based on butterfly and low-rank matrices, taking into account GPU block-oriented efficiency. Pixelated Butterfly trains up to 2.5x faster (wall-clock) than the dense Vision Transformer and GPT-2 counterparts with no drop in accuracy.
I will conclude by outlining future research directions for further accelerating ML pipelines and making ML more accessible to the general community, such as software-hardware co-design, data-centric AI, and ML for scientific computing and medical imaging.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Beidi Chen is a postdoctoral scholar in the CS department at Stanford University, working with Prof. Christopher Ré. Her research focuses on large-scale machine learning and deep learning. Specifically, she designs and optimizes randomized algorithms (algorithm-hardware co-design) to accelerate large machine learning systems for real-world problems. Prior to joining Stanford, she received her Ph.D. from the CS department at Rice University, advised by Prof. Anshumali Shrivastava. She received a BS in EECS from UC Berkeley. She has held internships in Microsoft Research, NVIDIA Research, and Amazon AI. Her work has won Best Paper awards at LISA and IISA. She was selected as a Rising Star in EECS by MIT and UIUC.
Host: Xiang Ren / Vatsal Sharan
Location: Ronald Tutor Hall of Engineering (RTH) - 105
Audiences: By invitation only.
Contact: Assistant to CS chair
-
CS Colloquium: Kuldeep Meel (National University of Singapore) - Counting, Sampling, and Synthesis: The Quest for Scalability
Mon, Apr 25, 2022 @ 09:00 AM - 10:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Kuldeep Meel, National University of Singapore
Talk Title: Counting, Sampling, and Synthesis: The Quest for Scalability
Abstract: The current generation of automated symbolic reasoning techniques excel at the qualitative tasks (i.e., when the answer is Yes
or No) owing to the dramatic progress in satisfiability solving, also referred to as the SAT revolution. The advances in SAT afford us the luxury to focus on quantitative reasoning tasks, whose development is critical to reason about the increasingly interconnected and complex computing systems.
In this talk, I will discuss the design of the next generation of automated reasoning techniques to perform higher-order tasks such as quantification (aka counting), sampling of representative behavior, and automated synthesis of systems. Naturally, these tasks are hard from a complexity-theoretic viewpoint, and therefore, our frameworks focus on tight integration of real-world applications, beyond the worst-case analysis algorithmic design and data-driven system design. This has allowed us to achieve significant advances in counting, sampling, and synthesis, providing a new algorithmic toolbox in formal methods, probabilistic reasoning, databases, and design verification. I will discuss the core design principles and the utility of the above techniques on various real applications, including quantitative analysis of AI systems and critical infrastructure resilience estimation.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Kuldeep Meel holds the NUS Presidential Young Professorship in the School of Computing at the National University of Singapore. His research interests lie at the intersection of Formal Methods and Artificial Intelligence. He is a recipient of the 2021 Amazon Research Award for Automated Reasoning, 2019 NRF Fellowship for AI, and was named AI's 10 to Watch by IEEE Intelligent Systems in 2020. His research program's recognition include the 2022 ACM SIGMOD Research Highlight, 2021 ICCAD Best Paper Award Nomination, "Best of PODS-21" invite from ACM TODS, "Best Papers of CAV-20" invite from FMSD journal, IJCAI-19 Sister conferences best paper award track invitation.
He holds a Ph.D. from Rice University, co-advised by Supratik Chakraborty and Moshe Y. Vardi. His thesis work received the 2018 Ralph Budd Award for Best Ph.D. Thesis in Engineering and the 2014 Outstanding Masters Thesis Award from Vienna Center of Logic and Algorithms, IBM PhD Fellowship, and Best Student Paper Award at CP 2015.
Host: Mukund Raghothaman
Webcast: https://usc.zoom.us/j/99187341067WebCast Link: https://usc.zoom.us/j/99187341067
Audiences: Everyone Is Invited
Contact: Cherie Carter
-
CS Distinguished Lecture: Henrik I. Christensen (UC San Diego) - Deploying autonomous vehicles for micro-mobility in urban environments
Tue, Apr 26, 2022 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Henrik I. Christensen, UC San Diego
Talk Title: Deploying autonomous vehicles for micro-mobility in urban environments
Series: Computer Science Distinguished Lecture Series
Abstract: Autonomous vehicles are already widely deployed on the inter-states. Providing robust autonomous systems for urban environments is a more difficult challenge, as the road network is more complex, there are many more types of road-users (cars, bikes, pedestrians) and the potential interactions are more complex. In an urban environment it is also harder to use pre-computed HD-maps as the world is more dynamic. We study the design of micro-mobility solutions for the UCSD campus. In this presentation we will discuss an overall systems design, eliminating the need for HD-maps and use course topological maps such as Open Street Maps, fusing vision and lidar for semantic mapping /localization, detection and handling other road-users. Dynamic planning in the presence of other agents. The system has been deployed in multiple long-term test to evaluate performance across weather, season changes, etc. We will present both underlying methods, algorithms, and experimental insights. Finally, we will present some challenges for the future.
Dr. Christensen will give the talk in person at SGM 124 and we will also host the talk over Zoom.
Register in advance for this webinar at:
https://usc.zoom.us/webinar/register/WN_M0LX4fKmSIqVhjLX05mWCg
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Henrik I Christensen is the Qualcomm Chancellor's Chair of Robot Systems and the director of robotics at UC San Diego. He is an academic, entrepreneur and investor. Dr. Christensen does research on a systems approach to robotics. Solutions need a solid theoretical basis, effective algorithms, a good implementation and must be evaluated using realistic scenarios. He has made contributions to computer vision, SLAM, and systems engineering. His research has been adopted by many companies. Henrik is also the main editor of the US National Robotics Roadmap (2009, 2013, 2016 and 2020). He is serving / has served on a significant number of editorial board (PAMI, IJRR, JFR, RAS, Aut Sys). He co-founded Robust.AI and Robo Global (AUM: $3.5B) and serves as a consultant to companies and agencies across 5 continents
Host: Stefanos Nikolaidis
Webcast: https://usc.zoom.us/webinar/register/WN_M0LX4fKmSIqVhjLX05mWCgLocation: Seeley G. Mudd Building (SGM) - 124
WebCast Link: https://usc.zoom.us/webinar/register/WN_M0LX4fKmSIqVhjLX05mWCg
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
CS Colloquium: Malte Jung (Cornell University) - Teamwork with Robots
Thu, Apr 28, 2022 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Malte Jung, Cornell University
Talk Title: Teamwork with Robots
Series: Computer Science Colloquium
Abstract: Research on Human-robot Interaction to date has largely focused on examining a single human interacting with a single robot. This work has led to advances in fundamental understanding about the psychology of human-robot interaction (e.g. how specific design choices affect interactions with and attitudes towards robots) and about the effective design of human-robot interaction (e,g. how novel mechanisms or computational tools can be used to improve HRI). However, the single-robot-single-human focus of this growing body of work stands in stark contrast to the complex social contexts in which robots are increasingly placed. While robots increasingly support teamwork across a wide range of settings covering search and rescue missions, minimally invasive surgeries, space exploration missions, or manufacturing, we have limited understanding of how groups people will interact with robots and how robots will affect how people interact with each other in groups and teams. In this talk I present empirical findings from several studies that show how robots can shape in direct but also subtle ways how people interact and collaborate with each other in teams.
Dr. Jung will give the talk in person at SGM 124 and we will also host the talk over Zoom.
Register in advance for this webinar at:
https://usc.zoom.us/webinar/register/WN_WnzW6E8UQQiLTd8N1tZoVw
After registering, attendees will receive a confirmation email containing information about joining the webinar.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Malte Jung is an Associate Professor in Information Science at Cornell University and the Nancy H. '62 and Philip M. '62 Young Sesquicentennial Faculty Fellow. His research brings together approaches from design and behavioral science to build understanding about how we can build robots that function better in group and team settings. His work has received several awards including an NSF CAREER award. He holds a Ph.D. in Mechanical Engineering, and a PhD Minor in Psychology from Stanford University. Prior to joining Cornell, Malte Jung completed a postdoc at the Center for Work, Technology, and Organization at Stanford University. He holds a Diploma in Mechanical Engineering from the Technical University of Munich.
Host: Stefanos Nikolaidis
Webcast: https://usc.zoom.us/webinar/register/WN_WnzW6E8UQQiLTd8N1tZoVwLocation: Seeley G. Mudd Building (SGM) - 124
WebCast Link: https://usc.zoom.us/webinar/register/WN_WnzW6E8UQQiLTd8N1tZoVw
Audiences: Everyone Is Invited
Contact: Computer Science Department
