Select a calendar:
Filter November Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for November
-
DREAM Industry Mentorship speaker series
Mon, Nov 04, 2024 @ 12:00 PM - 01:00 PM
USC Viterbi School of Engineering
University Calendar
DREAM connects students with experienced industry professionals from a variety of tech and destination companies who help them create a vision for their futures, align their careers around purpose, and build character in the context of growth, reinvention, and constant change. Industry mentors discuss how professional challenges present opportunities for character and leadership development. This event features strategic business executive Richard Au discussing the evolving landscape of big tech companies and finding work life balance.
Location: Ronald Tutor Hall of Engineering (RTH) - 217
Audiences: Everyone Is Invited
Contact: Elisabeth Arnold Weiss
Event Link: https://cglink.me/2nB/r400350
-
USC Quantum Technologies Forum
Thu, Nov 07, 2024 @ 09:00 AM - 05:30 PM
USC Viterbi School of Engineering
University Calendar
The USC Quantum Initiative will be publicly announced on Nov 7 with an all-day event (9-5:30) at USC Town and Gown Ballroom – USC Quantum Technologies Forum. We are expecting more than 150 attendees with more than 50 corporate representatives, including VCs and start-ups, plus USC quantum faculty, students, and staff, along with quantum faculty colleagues from other Southern California universities and some local govt officials. Must be registered by Nov 4 to attend. Contact: Maurena Nacheff-Benedict, Asst Dean, Viterbi Corporate & Foundation Relations
Location: Town & Gown (TGF) -
Audiences: By invitation
Contact: Andie Self
-
DREAM Industry Mentorship speaker series
Fri, Nov 08, 2024 @ 10:00 AM - 11:00 PM
USC Viterbi School of Engineering
University Calendar
DREAM connects students with experienced industry professionals from a variety of tech and destination companies who help them create a vision for their futures, align their careers around purpose, and build character in the context of growth, reinvention, and constant change. Industry mentors discuss how professional challenges present opportunities for character and leadership development. This event features digital media and entertainment leader Josh Auffret in conversation about his path from undergraduate film student at USC to head of UX program and operations at Google.
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: Everyone Is Invited
Contact: Elisabeth Arnold Weiss
Event Link: https://cglink.me/2nB/r400351
-
PhD Thesis Proposal - Xiao Fu
Wed, Nov 13, 2024 @ 02:00 PM - 03:30 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Computational Wildfire-proneLandscape Design and Mapping
Date and Time: Nov 13, 2 pm - 3:30 pm
Location: SAL 300
Committee members: Barath Raghavan, Bhaskar Krishnamachari, Ramesh Govindan, Peter Beerel, and Dani Yogatama
Abstract: Firefighters still rely on coarse remote sensing and inaccurate eyewitness reports to localize spreading wildfires. Despite advances in sensing, UAVs, and computer vision, the community has yet to combine the right modalities to achieve effective wildfire geolocalization and spotting. We present FireLoc, a fast and accurate wildfire crowdsensing system that localizes and maps wildfires combining ground cameras and landscape data. Prior image-based localization techniques fail in vegetated areas as they are tuned for close-range human-built environments. Instead, FireLoc integrates monocular depth mapping models, topography models, and cross-camera methods to achieve over 1000m range in vegetated environments leveraging low-cost smartphones. Due to the paucity of historical wildfire data, we built a wildfire simulator to provide additional data for validation. We show that FireLoc surpasses prior wildfire mapping work and reduces wildfire mapping time from hours to seconds.In future work, we propose a complete system that ensures landscape monitoring beyond the early wildfire propagation phase. We then emphasize multimodal approaches to landscape understanding for adaptive fuel analysis. Beyond monitoring the wildfire expansion, future systems can structurally understand the shifting landscape.Location: Henry Salvatori Computer Science Center (SAL) - 300
Audiences: Everyone Is Invited
Contact: Ellecia Williams
-
PhD Dissertation Defense - Ayush Jain
Wed, Nov 13, 2024 @ 04:30 PM - 06:30 PM
Thomas Lord Department of Computer Science
University Calendar
Presentation Title: Decision Making in Complex Action Spaces
Committee Members: Erdem Biyik, Joseph J Lim, Gaurav Sukhatme, Stefanos Nikolaidis, Feifei Quan
Date and Time: Wed., Nov. 13th, 2024: 4:30pm - 6:30pm
Location: VHE 214
Abstract: The action space of a reinforcement learning agent defines how it interacts with the world, whether selecting discrete items in a recommender system or controlling continuous movements in robotics. An agent is considered optimal when, at every step, it chooses an action within its action space that maximizes the expected future return. In this thesis, I study the relationship between expected returns and action space, identifying three key complexities that make certain tasks challenging. Specifically, I address decision-making in (1) unseen actions, such as new items to recommend; (2) changing action spaces, like variable inventory or toolset; and (3) locally optimal actions that hinder the search for the global best action. For each, we propose solutions to enhance agent adaptability and decision-making across complex action spaces.
Zoom Link: https://usc.zoom.us/j/91845196972?pwd=ghI6Q1gZmsmvonVUlFOTffDLAFwFY9.1Location: Vivian Hall of Engineering (VHE) - 214
Audiences: Everyone Is Invited
Contact: Ayush Jain
-
DREAM Industry Mentorship speaker series
Mon, Nov 18, 2024 @ 12:00 PM - 01:00 PM
USC Viterbi School of Engineering
University Calendar
DREAM connects students with experienced industry professionals from a variety of tech and destination companies who help them create a vision for their futures, align their careers around purpose, and build character in the context of growth, reinvention, and constant change. Industry mentors discuss how professional challenges present opportunities for character and leadership development. This event features actor and activist Emily Proctor on navigating life transitions and what she learned on her journey through Hollywood and beyond.
Location: Ronald Tutor Hall of Engineering (RTH) - 217
Audiences: Everyone Is Invited
Contact: Elisabeth Arnold Weiss
Event Link: https://cglink.me/2nB/r400352
-
PhD Thesis Proposal - Lixing Liu
Tue, Nov 19, 2024 @ 12:30 PM - 02:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Leveraging Organizational Hierarchies for Goal Management in Multi-Agent Reinforcement Learning
Date and Time: Nov. 19th, 2024 - 12:30p - 2:00p
Location: ICT 202
Committee Members: William Swartout (chair), Paul Rosenbloom, Kallirroi Georgila, Daniel O’Leary, Emilio Ferrara
Abstract: Effectively assigning credit and managing goals remain central challenges in Multi-Agent Reinforcement Learning (MARL), especially in stochastic environments with varying agent priorities across decision levels. Inspired by organizational hierarchies, this study structures multi-agent systems at different levels of abstraction and coordination. It hypothesizes that integrating a structured goal management mechanism within a MARL pipeline can: 1) improve the performance of prioritized, long-horizon tactical behaviors, 2) enhance the transferability of short-term operational behaviors, and 3) accelerate learning for faster MARL behavior model development. The proposed framework employs a hierarchy-aligned, soft-constraint goal-splitting strategy tailored to each agent’s capabilities, planning horizon, and organizational role. Furthermore, it enhances the manageability and interpretability of learned behaviors by incorporating sparse external graph networks to model environmental and inter-agent dynamics. This framework provides a solution for hierarchical goal management in MARL, evaluated for performance, efficiency and team coordination within two-team cooperative-competitive simulations involving complex maneuvers and engagements.
Zoom Link: https://usc.zoom.us/j/8634079147Location: Institute For Creative Technologies (ICT) - 202
Audiences: Everyone Is Invited
Contact: Lixing Liu
-
PhD Thesis Proposal - Hayley Song
Wed, Nov 20, 2024 @ 02:15 PM - 03:15 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Riemannian-Geometric Fingerprints of Generative Models
Date: November 20, 2024
Time: 2:15 pm - 3:15 pm
Location: KAP 209
Committee: Laurent Itti, Chair, Emilio Ferrara, Kyler Siegel, Robin Jia, and Willie Neiswanger
Abstract: Recent breakthroughs and rapid integration of generative models (GMs) have sparked interest in the problem of model attribution and their fingerprints.For instance, service providers need reliable methods of authenticating their models to protect their IP, while users and law enforcement seek to verify the source of generated content for accountability and trust. In addition, a growing threat of model collapse is arising, as more model-generated data are being fed back into sources (e.g., YouTube) that are often harvested for training ("regurgitative training''), heightening the need to differentiate synthetic from human data. Yet, a gap still exists in understanding generative models' fingerprints, we believe, stemming from the lack of a formal framework that can define, represent, and analyze the fingerprints in a principled way. To address this gap, we take a geometric approach and propose a new definition of artifact and fingerprint of generative models using Riemannian geometry, which allows us to leverage the rich theory of differential geometry.Our new definition generalizes previous work (Song et al, 2024) to non-Euclidean manifolds by learning Riemannian metrics from data and replacing the Euclidean distances and nearest-neighbor search with geodesic distances and kNN-based Riemannian center of mass. We apply our theory to a new gradient-based algorithm for computing the fingerprints in practice. Results show that it is more effective in distinguishing a large array of generative models, spanning across 4 different datasets in 2 different resolutions (64x64, 256x256), 27 model architectures, and 2 modalities (Vision, Vision-Language). Using our proposed definition can significantly improve the performance on model attribution, as well as a generalization to unseen datasets, model types, and modalities, suggesting its efficacy in practice.Location: Kaprielian Hall (KAP) - 209
Audiences: Everyone Is Invited
Contact: Ellecia Williams
-
PhD Thesis Defense - Pengmiao Zhang
Fri, Nov 22, 2024 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense: Pengmiao Zhang
Committee: Prof. Murali Annavaram, Prof. Rajgopal Kannan, Prof. Viktor K. Prasanna (Chair), Prof. Cauligi Raghavendra, Prof. Vatsal Sharan
Title:Machine Learning for Memory Access Prediction and Data Prefetching
Abstract:
Modern applications often experience performance bottlenecks due to memory system limitations. Data prefetching can hide memory access latency by predicting and loading data before it is needed. Machine Learning (ML) algorithms present a promising opportunity to enhance prefetching strategies. However, developing a high-performance ML-based prefetcher presents the following challenges: 1. ML modeling for memory access prediction, including extracting features from historical patterns, identifying future access targets, and designing models to capture their correlations. 2. Domain specific irregular memory access patterns due to multi-core execution and processing phases. 3. Balancing ML model complexity with hardware constraints, ensuring low-latency predictions while maintaining high performance. 4. Coordinated management of multiple prefetchers for ensemble prefetching. In this dissertation, we develop highly optimized ML models for data prefetching. First, to efficiently predict memory accesses for prefetching, we propose TransFetch, a novel attention-based approach that models prefetching as a multi-label classification problem. Second, we introduce a Domain Specific Machine Learning approach for prefetching, utilizing the context of architecture and computation to build high-performance ML-based prefetchers. Using this approach, we develop MPGraph and GraFetch to accelerate the execution of graph applications. Third, towards practical hardware deployment of ML-based prefetchers, we propose a novel tabularization approach that uses table hierarchies to approximate neural networks. We introduce DART, a table-based neural network prefetcher, and Net2Tab, a flexible tabularization framework. Lastly, we present ReSemble, an adaptive framework that uses reinforcement learning to optimize the coordination of multiple prefetchers. Our ML-based prefetchers show significant IPC improvements, demonstrating their performance advantages.
Bio: Pengmiao Zhang is a sixth-year PhD candidate in Computer Engineering, advised by Professor Viktor K. Prasanna. He received his BS degree in Electrical Engineering from Northeastern University (China) and MEng degree in Electrical Engineering from Harbin Institute of Technology. His research interests include machine learning for computer systems, memory system optimizations, and efficient machine learning.Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Julia Mittenberg-Beirao
Event Link: https://usc.zoom.us/j/9379439223
-
PhD Dissertation Defense - Jacqueline Brixey
Mon, Nov 25, 2024 @ 08:30 AM - 10:30 AM
Thomas Lord Department of Computer Science
University Calendar
Title: CODE-SWITCHING DIALOGUE SYSTEMS: AN INVESTIGATION INTO HOW SYSTEMS CAN SUPPORT CODE-SWITCHING AND WHEN THEY SHOULD, WITH ANALYSIS OF TWO CHOCTAW-ENGLISH APPLICATIONS
Date: November 25, 2024
Time: 8:30 am-10:30 am
Venue: USC ICT Room #202-Kilimanjaro
Committee: David Traum (chair), Maja Mataric, Khalil Iskarous
Abstract: This dissertation explores the development and application of bilingual dialogue systems, focusing specifically on systems that support English and Choctaw, an endangered American Indigenous language. Bilingual dialogue systems are critical in facilitating more natural and inclusive interactions for the many bilingual users worldwide, yet current systems often fail to accommodate linguistic features of bilingualism, such as code-switching.The dissertation investigates dialogue systems that manage unbalanced bilingualism and appropriate code-switching, improving user experience and system performance. I explore research questions such as whether code-switching leads to higher rapport, higher learning gains, or enhances interactions to collect endangered language audio data. Additionally, I address the sociocultural and linguistic challenges of developing conversational agents for endangered Indigenous languages.Location: USC ICT Room #202-Kilimanjaro
Audiences: Everyone Is Invited
Contact: Ellecia Williams
