Select a calendar:
Filter January Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for January
-
PhD Thesis Defense - Chung-Wei Lee
Wed, Jan 10, 2024 @ 01:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense - Chung-Wei Lee
Committee Members:
Haipeng Luo (chair)
Ashutosh Nayyar
Vatsal Sharan
Title:
No-Regret Learning and Last-Iterate Convergence in Games
Abstract:
No-regret learning (or online learning) is a general framework for studying sequential decision-making. Within this framework, the learner iteratively makes decisions, receives feedback, and adjusts their strategies. In this thesis, we consider analyzing the learning dynamics of no-regret algorithms in game scenarios where players play a single game repeatedly with particular no-regret algorithms. This exploration not only raises fundamental questions at the intersection of machine learning and game theory but also stands as a vital element when developing recent breakthroughs in artificial intelligence.
A notable instance of this influence is the widespread adoption of the “self-play” concept in game AI development, exemplified in games such as Go and Poker. With this technique, AI agents learn how to play by competing against themselves to enhance their performance step by step. In the terminology of literature focused on learning in games, the method involves running a set of online learning algorithms for players in the game to compute and approximate their game equilibria. To learn more efficiently in games, it is critical to design better online learning algorithms. Standard notions evaluating online learning algorithms in games include “regret,” assessing the average quality of iterates, and “last-iterate convergence,” representing the quality of the final iterates.
In this thesis, we design online learning algorithms and prove that they achieve near-optimal regret or fast last-iterate convergence in various game settings. We start from the simplest two-player zero-sum normal-form games and extend the results to multi-player games, extensive-form games that capture sequential interaction and imperfect information, and finally, the most general convex games. Moreover, we also analyze the weaknesses of prevalent online learning algorithms widely employed in practice and propose a fix for them. This not only makes the algorithms more robust but also sheds light on getting better learning algorithms for artificial intelligence in the future.Location: Ronald Tutor Hall of Engineering (RTH) - 306
Audiences: Everyone Is Invited
Contact: CS Events
-
PhD Thesis Defense - Meryem M'Hamdi
Tue, Jan 16, 2024 @ 10:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense - Meryem MHamdi
Title: Towards More Human-Like Cross-lingual Transfer Learning
Committee Members: Jonathan May (Chair), Aiichiro Nakano, Khalil Iskarous.
Abstract: Cross-lingual transfer learning comprises a set of techniques used to adapt a model trained on (a) source language(s), enabling it to generalize to new target languages. With the emergence of Transformer-based contextualized encoders, there has been a surge in multilingual representations that adapt these encoders to various cross-lingual downstream applications. The surprising zero-shot capabilities of these encoders make them promising substitutes for other fully supervised techniques, bypassing the need for large-scale annotation. However, these representations are still far from solving the long-tail of NLP phenomenon, where models are biased more towards high-resource and typologically similar languages to the source language. This bias can be attributed to the over-reliance of current transfer learning pipelines on what we define as the 'Data-Intensive Identically-Distributed Minimally-Evaluated' paradigm. In this thesis, we analyze and propose techniques to advance the capabilities of multilingual language models beyond the traditional paradigm and more toward human-like cross-lingual transfer learning. We achieve that through 1) human-inspired input requirements by using few-shot meta-learning techniques, 2) human-inspired outcomes by defining a cross-lingual continual learning evaluation paradigm, and 3) human-inspired approaches through devising cognitive strategies to consolidate retention of knowledge learned across languages. Our contributions towards advancing the current transfer learning paradigm towards human-like learning are four-fold: 1) We explore cross-lingual fine-tuning on low-resource multilingual applications such as event trigger extraction and semantic search, shedding light on the strengths and limitations of existing cross-lingual transfer learning techniques. 2) We propose language-agnostic meta-learning approaches that can further bridge the gap between source and target typologically diverse languages. We show the merits of our approaches in reaching quicker and smoother generalization compared to naive fine-tuning, especially under low-resource scenarios. 3) We are the first to define a lifelong learning paradigm that analyzes language shifts. We show the merits and challenges of a multi-hop analysis where the system continually learns over several languages one at a time. 4) We are the first to adapt a cognitively inspired technique based on Leitner-queues to choose what to repeat in a cross-lingual continual learning setup and investigate its impact on reducing the forgetting of previously learned languages.Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: CS Events
Event Link: https://usc.zoom.us/j/94477374759?pwd=ajZrd1o0QktVVXZsRk9UL1J6NGdtdz09#success
-
PhD Dissertation Defense - Xuefeng Hu
Fri, Jan 19, 2024 @ 03:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Dissertation Defense - Xuefeng Hu
Committee members: Ram Nevatia (chair), Aram Galstyan and Keith Jenkins
Title: Adapt Pre-trained Representation Towards Downstream Tasks
Abstract: In recent years, the field of computer vision and machine learning has witnessed a paradigm shift, characterized by a dramatic increase in the scale of model parameters and training data sizes. This evolution has led to significant enhancements in model accuracy and robustness, transcending the traditional, task-specific expert models. The field has now pivoted towards universal, large-scale pre-trained visual representations, which enables impressive zero-shot and few-shot solutions for a wide array of downstream tasks.
Despite these advancements, the application of pre-trained models to specific downstream tasks, each with their unique conditions and domain-specific challenges, often exposes inherent limitations. This dissertation aims to tackle these challenges. The research journey comprises a spectrum of approaches from fully-supervised to source-free and test-time adaptation, with diverse applications such as image classification, object detection, and forensic detection. This dissertation introduces novel architectures such as SPAN, which has pioneered the utilization of the self-attention mechanism in the field of computer vision, as well as innovative adaptation algorithms like ReCLIP and BaFTA, which enhance zero-shot classification performance with unsupervised vision-text alignment. This dissertation marks a transition from classic visual representations, like those used in ImageNet, to cutting-edge vision-language models like CLIP, and has overcome some of the most pressing challenges in the field.
The works of this dissertation play an important role in bridging the gap between generic visual representations and the specific, nuanced requirements of various real-world tasks. By doing so, it establishes new benchmarks in optimizing the performance of machine learning models in practical applications, reinforcing the role of advanced computational techniques in solving complex, real-world problems.
Zoom Link: https://usc.zoom.us/j/95935934090?pwd=RTFNcUorbndkaXA2UGtFWWkrbEtsUT09
Meeting ID: 959 3593 4090
Passcode: 442518
Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: CS Events
Event Link: https://usc.zoom.us/j/95935934090?pwd=RTFNcUorbndkaXA2UGtFWWkrbEtsUT09
-
PhD Thesis Defense - Jiao Sun
Wed, Jan 24, 2024 @ 11:00 AM - 01:30 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense - Jiao Sun
Committee members: Xuezhe (Max) Ma (Chair), Nanyun (Violet) Peng, Johnathan May, Emilio Ferrara, Dan O’Leary
Title: Emphasizing the Importance of Data and Evaluation in the Era of Large language Models
Abstract: Large-scale models have marked the beginning of a new era, significantly transforming language understanding, text and image generation, and complex decision-making tasks. For example, they may perpetuate stereotypes or produce misleading information. Nevertheless, due to the limitations of existing evaluation methods, these problems are often overlooked. My research highlights the imperative need for more careful and nuanced model evaluation and assessment. Upon investigation, large models may have generated biased or inappropriate content due to inadequacies in their training data. Identified through trustworthy evaluation methods, I address these challenges with a focus on aligning large models with human intentions from a data-centric perspective.Location: Ronald Tutor Hall of Engineering (RTH) - 306
Audiences: Everyone Is Invited
Contact: CS Events
Event Link: https://usc.zoom.us/my/jiaosun
-
PhD Thesis Proposal - Haowen Lin
Wed, Jan 24, 2024 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
Committee: Cyrus Shahabi (Chair), Bistra Dilkina, Muhao Chen, Xiong Li, Marlon Boarnet
Title: Accurate and controllable trajectory generation
Abstract : In various application domains like transportation, urban planning, and public health, analyzing human mobility, represented as a sequence of consecutive visits (aka trajectories), is crucial for uncovering essential mobility patterns. Due to privacy and commercial concerns, real-world trajectories are not readily available, giving rise to an important research area of generating synthetic but realistic trajectories. This thesis addresses the challenge of trajectory generation using data-driven approaches, integrating both explicit and implicit constraints within a continuous spatiotemporal domain. First, I present a framework based on generative adversarial imitation learning that synthesizes realistic trajectories that preserve moving behavior patterns (.g., work commute, shopping purpose) in real data. Next, I explore the hypothesis that grouping trajectories governed by similar dynamics into clusters before trajectory modeling could enhance modeling effectiveness. I present a framework that can simultaneously model trajectories in continuous space and time while clustering them. Finally, we discuss the proposed work that will incorporate explicit spatial and temporal constraints that will potentially generate more representative and realistic trajectories.
Zoom link: https://usc.zoom.us/j/95828555243Location: https://usc.zoom.us/j/95828555243
Audiences: Everyone Is Invited
Contact: CS Events
Event Link: zoom link: https://usc.zoom.us/j/95828555243
