Select a calendar:
Filter May Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for May 10, 2021
-
Department of Biomedical Engineering Seminar - Dr. Kristin Swanson
Mon, May 10, 2021 @ 09:00 AM - 10:00 AM
Alfred E. Mann Department of Biomedical Engineering
Conferences, Lectures, & Seminars
Speaker: Dr. Kristin Swanson , Professor & Vice Chair of Research, Neurological Surgery in Arizona, Mayo Clinic Professor, Mathematics, Arizona State University
Talk Title: Every Patient Deserves Their Own Equation
Host: Kirk Shung
More Information: Dr. Kristin Swanson Flier as of 4 26.pdf
Audiences: Everyone Is Invited
Contact: Michele Medina
-
PhD Defense - Nitin Kamra
Mon, May 10, 2021 @ 10:30 AM - 12:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Candidate: Nitin Kamra
Committee: Prof. Yan Liu (chair), Prof. Bistra Dilkina, Prof. Ashutosh Nayyar
Date: 10th May, 2021
Time: 10:30am-12:00pm
Zoom: https://usc.zoom.us/j/96990524233?pwd=M3BzcTdOenZtRjlkN1J5dmxDQmVpUT09
Meeting ID: 969 9052 4233
Passcode: 015551
Title: Machine Learning in Interacting Multi-agent Systems
Abstract:
Making predictions and learning optimal behavioral strategies are important problems in many domains such as traffic prediction, pedestrian tracking, financial investments and security systems. These systems often consist of multiple agents interacting with each other in complex ways, which makes both the above tasks very challenging. In this thesis, I propose methods to advance the state-of-the-art for such multi-agent learning problems. The first part of my talk focuses on trajectory prediction and I will present a relational model involving a fuzzy decision making attention mechanism for multi-agent trajectory prediction. Our approach shows significant performance gains over many existing state-of-the-art predictive models in diverse domains such as human crowds, US freeway traffic and various physics datasets. The second part of my talk focuses on placing multiple resources to protect and cover geographical spaces. We propose the Coverage Gradient Theorem and a spatial discretization based framework to improve existing benchmarks for spatial coverage domains. The third part of the talk focuses on computing nash equilibrium strategies in spatial security games with continuous action spaces. We present our model-free learning algorithm, OptGradFP, and our model-based learning algorithm, DeepFP, which search for the optimal defender strategy in a parameterized continuous search space. These algorithms scale to large domains and compute strategies robust to adversarial exploitation. Finally, we combine the Coverage Gradient framework with DeepFP to show improved performance on spatial coverage security domains.
WebCast Link: https://usc.zoom.us/j/96990524233?pwd=M3BzcTdOenZtRjlkN1J5dmxDQmVpUT09
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
Thesis Proposal - Shen Yan
Mon, May 10, 2021 @ 11:00 AM - 01:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Fair Machine Learning for Human Behavior Understanding
Time: 11:00 AM-1:00 PM PST, May 10 (Monday)
Committee: Emilio Ferrara, Cyrus Shahabi, Shri Narayanan, Kristina Lerman, and Fred Morstatter.
Zoom link: https://usc.zoom.us/j/96050343860
Abstract:
Artificial intelligence (AI) and machine learning models have been recently applied extensively to understand and predict human behavior, often in applications with major societal implications, such as making recruitment decisions, estimating daily well-beings, or assessing clinical treatments. Despite the increasing body of research on modeling human behavior and fair machine learning, most studies focus on homogeneous and objective measurements, and little has been discussed on how to mitigate the impact of heterogeneity on utility and fairness simultaneously. The increasing amount of collected data also raises concerns on data privacy. Recent regulations such as European General Data Protection Regulation (GDPR) and California Consumer Privacy Act (CCPA) regulate the usage of personal data. However, most previous fairness work requires the access of sensitive attributes (e.g., race, gender) to debias the system.
This dissertation proposal will articulate the challenges posed by complex, multimodal human behavior data in both model utility and fairness. The proposed work is decomposed into three tasks, namely tackling machine learning fairness issues originating from the heterogeneous human behaviors (Task 1), and biased behavior annotations (Task 2), and designing fair machine learning methods without sensitive attributes (Task 3) for both centralized and federated learning. This work will provide possible solutions to mitigate bias of human behavior understanding systems, reducing barriers to access, alleviating systemic racism, discrimination, and unfair process.WebCast Link: https://usc.zoom.us/j/96050343860
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
Thesis Proposal - Jiaoyang Li
Mon, May 10, 2021 @ 02:00 PM - 03:30 PM
Thomas Lord Department of Computer Science
University Calendar
Title:
Efficient and Effective Multi-Agent Path Finding via Heuristic Search, Symmetry Breaking, and Large Neighborhood Search
Time:
2:00-3:30pm PST, May 10 (Monday)
Committee: Sven Koenig, Nora Ayanian, Bistra Dilkina, T. K. Satish Kumar, Satyandra K. Gupta, and Brian Williams (MIT)
Zoom link:
https://usc.zoom.us/j/96942890548?pwd=SkFzQm80cHNOMHBPQXZBa1N3eVZvZz09
Abstract:
Recent advances in robotics have laid the foundation for building large-scale multi-agent systems. One fundamental task in many multi-agent systems is to navigate teams of agents in a shared environment to their target locations without colliding with obstacles or other agents. Applications include evacuation, formation control, object transportation, traffic management, search and rescue, autonomous driving, drone swarm coordination, video game character control, and large-scale warehouse automation, to list a few. One well-studied abstract model for this problem is known as Multi-Agent Path Finding (MAPF). MAPF is NP-hard to solve optimally (or, in some cases, even bounded-suboptimally) in general. Existing algorithms for solving MAPF either have limited scalability, generate costly solutions, or fail to find any solutions for hard MAPF problems. I propose to develop fundamental techniques for solving MAPF more efficiently and effectively that exploit the combinatorial structure of the MAPF problem and combine ideas from both AI and OR. In particular, I design admissible heuristics by cost partitioning (ideas from the AI planning community) to speed up optimal MAPF algorithms. I also develop symmetry-breaking constraints (ideas from the constraint programming community) to eliminate symmetries in optimal MAPF solving. For future work, I plan to learn inadmissible heuristics (ideas from the heuristic search community) to speed up bounded-suboptimal MAPF algorithms and build an anytime MAPF framework via large neighborhood search (ideas from the constraint programming and operations research communities) to improve the success rate and the solution quality of non-optimal MAPF algorithms.WebCast Link: https://usc.zoom.us/j/96942890548?pwd=SkFzQm80cHNOMHBPQXZBa1N3eVZvZz09
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
