Select a calendar:
Filter November Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for November 30, 2021
-
Differential Verification of Deep Neural Networks
Tue, Nov 30, 2021 @ 08:30 AM - 09:30 AM
Thomas Lord Department of Computer Science
Student Activity
PhD Candidate: Brandon Paulsen
Title: Differential Verification of Deep Neural Networks
Date & Time: Tuesday November 30th at 8:30 AM
Committee: Chao Wang (Advisor), Jyotirmoy Deshmukh, Nenad Medvidovic, William Halfond, Murali Annavaram
Zoom link: https://usc.zoom.us/j/97339789019?pwd=blJoYTg3WXJDZzBUcFVRQzZMNUNpQT09
Abstract:
Recently, deep neural networks (DNNs) have found success in a wide variety of application domains such as image recognition, natural language processing, and autonomous vehicle control. However, they are often criticized for their large energy-footprint, which limits their use on computationally- and energy-constrained devices. Recently, this limitation was addressed by DNN compression -- a technique that reduces the computational and energy requirements by, e.g., reducing the floating point precision of the neural network -- but this naturally raises the question: is the compressed network equivalent to the original? Answering this question is crucial for safety-critical systems, and desirable in general. Unfortunately, current DNN verification tools are limited in that they are only designed to analyze a single network, rendering them ineffective for this problem.
For my thesis, I address this limitation by formalizing the problem of differential verification of DNNs, and then developing a novel approach for reasoning about a pair of any two structurally similar feed-forward DNNs with ReLU activations. The key insight in my approach is to reason about the two networks simultaneously, thus greatly improving the precision of the analysis. While the approach is applicable to any pair of structurally similar DNNs, I demonstrate its effectiveness in proving equivalence (within a small error bound) of compressed DNNs with respect to the original DNN, and I further show that my new approach outperforms existing DNN verification tools by orders of magnitude, in terms of scalability. I then show that the first approach can be greatly improved upon by leveraging a novel fine-grained, symbolic technique that captures the relationships between neurons. Finally, I discuss the challenges of extending differential verification to activation functions beyond ReLU and other DNN architectures, and propose a solution.
Location: Zoom
WebCast Link: https://usc.zoom.us/j/97339789019?pwd=blJoYTg3WXJDZzBUcFVRQzZMNUNpQT09
Audiences: Everyone Is Invited
Contact: USC Computer Science
-
Astani Civil and Environmental Engineering Seminar
Tue, Nov 30, 2021 @ 11:00 AM - 12:00 PM
Sonny Astani Department of Civil and Environmental Engineering
Conferences, Lectures, & Seminars
Speaker: Adam Simpson, Ph.D. Candidate, Stanford University
Talk Title: Predicting Initial Transformation of Food-Based Bio-Polymers and -Molecules During Food Disinfection
Abstract: Foodborne pathogenic outbreaks still occur 100x more frequently than waterborne pathogenic outbreaks from municipal/community water supplies in the US, despite the passing of the Food Safety Modernization Act of 2011. To kill pathogens, food packaging facilities must treat fruits, vegetables, and meats with high dosages of chemical sanitizers-”most frequently free chlorine-”for ready-to-eat and triple washed foods. Food chemists and engineers have looked to water disinfection byproduct research to predict and measure contaminants (i.e., trihalomethanes, haloacetic acids and chlorates) formed in food disinfection washwaters as potential chemical exposures to consumers. However, my research demonstrates that this is fundamentally incorrect as the boundary conditions of food disinfection and water disinfection processes
are incongruent. First, I will introduce current food disinfection processing techniques and conditions and compare with water disinfection to encourage a new way to predict initial transformation products formed during food disinfection. Second, I will demonstrate the formation of chlorotyrosines isolated inside of chlorine washed spinach and lettuce in comparison to volatile disinfection byproducts isolated in the washwater to prove where true toxic exposure risks are. Third, I will introduce a new class of food-based disinfection byproducts measured inside of chlorine treated vegetables, namely fatty acid chlorohydrins,and compare the cumulative toxic potencies of these chlorine treated vegetables to a poor-quality water at EPA regulated limits to demonstrate the relative magnitude of these exposures. I will end the seminar with a glance of my future research involving this uncharted field.
Biography: Adam Simpson is a PhD candidate at Stanford University, in the Civil and Environmental Engineering Department advised by Professor William Mitch. His research involves using organic chemistry and complex analytical chemistry techniques to synthesize and isolate new classes of food-based disinfection
byproducts that are toxic and affect most consumers. To advance his depth of knowledge and experience with postharvest crop treatment, he is also undergoing a stint at the United States Department of Agriculture-”Agricultural Research Service in Parlier, California, where he is also advised by the
Environmental Chemist, Dr. Spencer Walse. His work has been supported by multiple fellowships including a Stanford Graduate Fellowship, a National Science Foundation Graduate Research Fellowship, and a Diversifying Academia, Recruiting Excellence Fellowship. He will receive his PhD in Civil and Environmental
Engineering in May-2022 and received his M.S. degree in Civil and Environmental Engineering from Stanford University, and B.S. degree in Chemical Engineering from Carnegie Mellon University. Beingpassionate about diversifying academia, Adam is also a science communicator through his YouTube Channel, where he aims to humanize academia with a focus on Science, Technology, Engineering, and Math Education, while uplifting underrepresented voices. With an ambition of being a tenure-track assistant
professor of Civil and Environmental Engineering-”he hopes to continue his work to make a more inclusive
academia.
Host: Dr. Daniel McCurry
Webcast: https://usc.zoom.us/j/94682875854 Meeting ID: 946 8287 5854 Passcode: 090068Location: Michelson Center for Convergent Bioscience (MCB) - 102
WebCast Link: https://usc.zoom.us/j/94682875854 Meeting ID: 946 8287 5854 Passcode: 090068
Audiences: Everyone Is Invited
Contact: Evangeline Reyes
-
Photonics Seminar Series
Tue, Nov 30, 2021 @ 01:30 PM - 02:30 PM
Ming Hsieh Department of Electrical and Computer Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Juliet Gopinath, University of Colorado Boulder
Talk Title: Nonlinear Integrated Optics and Microscopy
Series: Photonics Seminar
Host: Electrical and Computer Engineering: Wade Hsu, Mercedeh Khajavikhan, Michelle Povinelli, Constantine Sideris, and Wei Wu
More Info: https://usc.zoom.us/j/91808071892?pwd=VUwyK3NSNW5rSzVLQzFKSGdPc05yUT09
Webcast: https://usc.zoom.us/j/91808071892?pwd=VUwyK3NSNW5rSzVLQzFKSGdPc05yUT09More Information: Photonics Seminar _Julie Gopinath 11-30-21.pdf
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132
WebCast Link: https://usc.zoom.us/j/91808071892?pwd=VUwyK3NSNW5rSzVLQzFKSGdPc05yUT09
Audiences: Everyone Is Invited
Contact: Jennifer Ramos/Electrophysics
Event Link: https://usc.zoom.us/j/91808071892?pwd=VUwyK3NSNW5rSzVLQzFKSGdPc05yUT09
-
Drop-In Weekly Office Hours [Virtual] Posted By: Center for Advanced Research ComputingTue, Nov 30, 2021 @ 02:30 PM - 05:00 PM
Information Technology Program (ITP)
Workshops & Infosessions
Every Tuesday, office hours are an opportunity for CARC users to ask questions about research computing. No appointment/registration is necessary, but you must use your USC credentials to access the Zoom meeting by clicking "Register" below. For in-person support, we are also in Leavey Library room 3M (basement) during this same time period. Register Here!
Audiences: Everyone Is Invited
Contact: Center for Advanced Research Computing
-
ISE 651 - Epstein Seminar
Tue, Nov 30, 2021 @ 03:30 PM - 04:50 PM
Daniel J. Epstein Department of Industrial and Systems Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Xunyu Zhou, Professor, Dept. of Industrial Engineering and Operations Research, Columbia
Talk Title: Policy Evaluation, Policy Gradient, and Actor-Critic Learning in Continuous Time and Space: Theory and Algorithms
Host: Dr. Renyuan Xu
More Information: November 30, 2021.pdf
Location: Zoom/Online
Audiences: Everyone Is Invited
Contact: Grace Owh
-
CS Colloquium: Konstantinos Karydis (University of California, Riverside) - Online mobile robot motion planning under uncertainty in unknown environments
Tue, Nov 30, 2021 @ 03:30 PM - 04:50 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Konstantinos Karydis, University of California, Riverside
Talk Title: Online mobile robot motion planning under uncertainty in unknown environments
Series: Computer Science Colloquium
Abstract: Mobile robot motion planning under uncertainty is a challenging yet rewarding foundational robotics research problem with extensive applications across domains including intelligence, surveillance and reconnaissance (ISR), remote sensing, and precision agriculture. One important challenge is operation in unknown environments where planning decisions need to be made at run-time. In this talk we discuss recent results to address online motion planning in unknown environments. We consider two specific cases: 1) How to achieve resolution-complete field coverage considering the non-holonomic mobility constraints in commonly-used vehicles (e.g., wheeled robots) without prior information about the environment? 2) How to develop resilient, risk-aware and collision-inclusive planning algorithms to enable (collision-resilient) mobile robots to deliberately choose when to collide with locally-sensed obstacles to improve some motion planning metrics (e.g., total time to reach a goal).
To this end, we have proposed a hierarchical, hex-decomposition-based coverage planning algorithm for unknown, obstacle-cluttered environments. The proposed approach ensures resolution-complete coverage, can be tuned to achieve fast exploration, and plans smooth paths for Dubins vehicles to follow at constant velocity in real-time. Our approach can successfully trade-off between coverage and exploration speed, and can outperform existing online coverage algorithms in terms of total covered area or exploration speed according to how it is tuned. Further, we have introduced new sampling- and search-based online collision-inclusive motion planning algorithms for impact-resilient robots, that can explicitly handle the risk of colliding with the environment and can switch between collision avoidance and collision exploitation. Central to the planners' capabilities is a novel joint optimization function that evaluates the effect of possible collisions using a reflection model.
This way, the planner can make deliberate decisions to collide with the environment if such collision is expected to help the robot make progress toward its goal. To make the algorithm online, we present state expansion pruning techniques that can significantly reduce the search space while ensuring completeness.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Dr. Karydis is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of California, Riverside (UCR). Before joining UCR, he worked as a Post-Doctoral Researcher in Robotics in GRASP Lab, which is part of the Department of Mechanical Engineering and Applied Mechanics at the University of Pennsylvania (Penn). His work was supported by Dr. Vijay Kumar, Professor and Nemirovsky Family Dean of Penn Engineering. He completed his doctoral studies in the Department of Mechanical Engineering at the University of Delaware, under the guidance of Prof. Herbert Tanner and Prof.
Ioannis Poulakakis.
Host: Stefanos Nikolaidis
Location: Seeley Wintersmith Mudd Memorial Hall (of Philosophy) (MHP) - 101
Audiences: Everyone Is Invited
Contact: Computer Science Department
-
Virtual First-Year Admission Information Session
Tue, Nov 30, 2021 @ 04:00 PM - 05:00 PM
Viterbi School of Engineering Undergraduate Admission
Workshops & Infosessions
Our virtual information session is a live presentation from a USC Viterbi admission counselor designed for high school students and their family members to learn more about the USC Viterbi undergraduate experience. Our session will cover an overview of our undergraduate engineering programs, the application process, and more on student life. Guests will be able to ask questions and engage in further discussion toward the end of the session.
Register here!Audiences: Everyone Is Invited
Contact: Viterbi Admission
