Select a calendar:
Filter April Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for April
-
PhD Defense - Jared Coleman
Thu, Apr 04, 2024 @ 10:00 AM - 11:30 AM
Thomas Lord Department of Computer Science
University Calendar
PhD Defense: Jared Coleman
Title: Dispersed Computing for Dynamic Environments Committee: Bhaskar Krishnamachari (Chair), Konstantinos Psounis, Jyotirmoy Deshmukh
Abstract: Scheduling a distributed application modeled as a directed acyclic task graph over a set of networked compute nodes is a fundamental problem in distributed computing and thus has received substantial scholarly attention. Most existing solutions, however, fall short of accommodating the dynamic and stochastic nature of modern dispersed computing systems (e.g., IoT, edge, and robotic systems) where applications and compute networks have stricter and less stable resource constraints. In this dissertation, we identify problems and propose solutions that address this gap and advance the current state-of-the-art in task scheduling.Location: Ronald Tutor Hall of Engineering (RTH) - 211
Audiences: Everyone Is Invited
Contact: Asiroh Cham
-
PhD Thesis Proposal - Qinyuan Ye
Mon, Apr 22, 2024 @ 10:00 AM - 11:30 AM
Thomas Lord Department of Computer Science
University Calendar
Title: Cross-Task Generalization Abilities of Large Language Models
Committee Members: Xiang Ren (Chair), Robin Jia, Swabha Swayamdipta, Jesse Thomason, Morteza Dehghani
Date & Time: Monday, April 22, 10am-11:30am\
Location: SAL 213
Abstract: Humans can learn a new language task efficiently with only a few examples, by leveraging their knowledge and experience obtained when learning prior tasks. Enabling similar cross-task generalization abilities in NLP systems is fundamental for achieving the goal of general intelligence and enabling broader and more scalable adoption of language technology in future applications. In this thesis proposal, I will present my work on (1) benchmarking cross-task generalization abilities with diverse NLP tasks; (2) developing new model architecture for improving cross-task generalization abilities; (3) analyzing and predicting the generalization landscape of current state-of-the-art large language models. Additionally, I will outline future research directions, along with preliminary thoughts on addressing them.
Zoom Link: https://usc.zoom.us/j/93269270403?pwd=NVNmN085bm5SWXNnNGErcXczeVkxdz09Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: Qinyuan Ye
Event Link: https://usc.zoom.us/j/93269270403?pwd=NVNmN085bm5SWXNnNGErcXczeVkxdz09
-
PhD Dissertation Defense - Arka Sadhu
Tue, Apr 23, 2024 @ 02:00 PM - 03:30 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Grounding Language in Images and Videos
Location: SAL 213
Time: 2 pm on April 23, 2024
Committee Members: Ram Nevatia (Chair), Xiang Ren, Toby Mintz
Abstract: My thesis investigates the problem of grounding language in images and videos -- the task of associating linguistic symbols to perceptual experiences and actions -- which is fundamental to developing multi-modal models that can understand and jointly reason over images, videos, and text. The overarching goal of my dissertation is to bridge the gap between language and vision as a means to a ``deeper understanding'' of images and videos to allow developing models capable of reasoning over longer-time horizons such as hour-long movies, or a collection of images, or even multiple videos. In this thesis, I will introduce the various vision-language tasks developed during my Ph.D. which include grounding unseen words, spatiotemporal localization of entities in a video, video question-answering, and visual semantic role labeling in videos, reasoning across more than one image or a video, and finally, weakly-supervised open-vocabulary object detection. For each of these tasks, I will further discuss the development of corresponding datasets, evaluation protocols, and model frameworks. These tasks aim to investigate a particular phenomenon inherent in image or video understanding in isolation, develop corresponding datasets and model frameworks, and outline evaluation protocols robust to data priors.
The resulting models can be used for other downstream tasks like obtaining common-sense knowledge graphs from instructional videos or drive end-user applications like Retrieval, Question Answering, and Captioning.
Zoom Link: https://usc.zoom.us/j/94652316277?pwd=QTdqcklJMjg2UE03ZVZHbmFvWU9nQT09Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: Arka Sadhu
-
PhD Thesis Defense - Pei Zhou
Wed, Apr 24, 2024 @ 02:00 AM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense - Pei Zhou
Committee Members: Xiang Ren (Chair), Jay Pujara (Co-Chair), Toby Mintz, Jieyu Zhao
Title: Common Ground Reasoning for Communicative Agents
Abstract: Effective communication requires reasoning to reach mutual beliefs and knowledge among participants, a process called grounding. Large language model (LLM)-powered conversational AIs have displayed impressive capabilities, showing the potential of building AI agents that can interact with humans and the world smoothly. However, challenges remain unsolved for AI models to become capable communicative agents including understanding implicit intents and reaching goals. My PhD thesis outlines my research aiming to tackle these challenges by teaching models to reason to build common ground to become better communicators. Specifically, I focus on 1) enhancing conversational models with common sense knowledge; 2) modeling theory-of-mind capabilities to build goal-driven dialogue agents; and 3) eliciting metacognition by planning reasoning strategies for diverse scenarios. I will also discuss future directions including life-long self-learning with evolving common ground for personalization, interactive super-alignment to supervise models stronger than us, and measuring and improving safety to deploy agents in the wild.
Venue: RTH 306 and Zoom https://usc.zoom.us/j/2065614640
Date: 04/24/2024, 2-4PMLocation: Ronald Tutor Hall of Engineering (RTH) -
Audiences: Everyone Is Invited
Contact: CS Events
Event Link: https://usc.zoom.us/j/2065614640
-
PhD Thesis Proposal - Navid Hashemi
Thu, Apr 25, 2024 @ 10:30 AM - 12:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Verification and Synthesis of Controllers for Temporal Logic Objectives Using Neuro-Symbolic Methods
Committee Members: Jyotirmoy Deshmukh (Chair), Guarav Sukhatme, Chao Wang, Pierlggi Nuzzo, Lars Lindemann, Georgios Fainekos (External Member)
Date & Time: Thursday, April 25th, 10:30am - 12:00pm
Abstract: As the field of autonomy is embracing the use of neural networks for perception and control, Signal Temporal Logic (STL) has emerged as a popular formalism for specifying the task objectives and safety properties of such autonomous cyber-physical systems (ACPS). There are two important open problems in this research area: (1) how can we effectively train neural controllers in such ACPS applications, when the state dimensionality is high and when the task objectives are specified over long time horizons, and (2) how can we verify if the closed-loop system with a given neural controller satisfies given STL objectives. We review completed work in which we show how discrete-time STL (DT-STL) specifications lend themselves to a smooth neuro-symbolic encoding that enables the use of gradient-based methods for control design. We also show how a type of neuro-symbolic encoding of DT-STL specifications can be combined with neural network verification tools to provide deterministic guarantees. We also review how neural network encoding of the environment dynamics can help us combine statistical verification techniques with formal techniques for reachability analysis. We will then propose several directions that we will pursue in the future: (1) We will investigate if our neuro-symbolic encoding approach can extend to other temporal logics, especially those used for specifying properties of perception algorithms (such as Spatio-Temporal Perception Logic or STPL). Our idea is to use a neuro-symbolic encoding of STPL to improve the quality of outputs produced by perception algorithms. (2) We will investigate how control policies generated by our existing algorithms can be made robust to distribution shifts through online and offline techniques. (3) Finally, we will propose scaling our synthesis approaches to higher-dimensional observation spaces and longer horzon tasks. We conclude with the timeline to finish proposed work and write the dissertation.Location: Ronald Tutor Hall of Engineering (RTH) - 306
Audiences: Everyone Is Invited
Contact: Felante' Charlemagne
-
Phd Dissertation Defence - Haidong Zhu
Thu, Apr 25, 2024 @ 12:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Shape-Assisted Multimodal Person Re-Identification
Committee Members: Ram Nevatia (Chair), Ulrich Neumann, Antonio Ortega
Date & Time: Thursday, April 25th, 12:00pm - 2:00pm
Abstract: Recognizing an individual's identity across non-overlapping images or videos, known as person re-identification, is a fundamental yet challenging task for biometric analysis. This task involves extracting and distinguishing unique features such as appearance, gait, and body shape to accurately identify individuals. Different from other representations, 3-D shape complements the body information with external human body shape prior and enhances the appearance captured in the 2-D images. Although 3-D body shape offers invaluable external shape-related information that 2-D images lack, existing body shape representations often fall short in accuracy or demand extensive image data, which is unavailable for re-identification tasks. We explore various biometric representations for comprehensive whole-body person re-identification, with a particular emphasis on leveraging 3-D body shape. We focus on enhancing the detail and few-shot learning capabilities of 3-D shape representations through the application of implicit functions and generalizable Neural Radiance Fields (NeRF). Moreover, we propose the use of 3-D body shape for alignment and supervision during training, aiming to advance the accuracy and efficiency of person re-identification techniques.Location: Hughes Aircraft Electrical Engineering Center (EEB) - 110
Audiences: Everyone Is Invited
Contact: Haidong Zhu
-
PhD Dissertation Defense - Zhaoheng Zheng
Thu, Apr 25, 2024 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Incorporating Large-Scale Vision-Language Corpora in Visual Understanding
Committee Members: Ram Nevatia (Chair), Mohammad Soleymani, Keith Jenkins
Date and Time: Thursday, April 25th, 2:00pm - 4:00pm
Abstract: As key mediators of human perception, vision and language corpora act as critical roles in the development of modern Artificial Intelligence (AI). The size of vision-language corpora has scaled up rapidly in recent years, from thousands to billions, enabling the creation of large foundation models. However, as an emerging concept, there are a series of problems yet to be explored.
We start with a study of compositional learning from pre-VLM times to the post-VLM era. We introduce a representation blending approach that creates robust features for compositional image classification and a two-stream architecture that tackles the entanglement in the feature space of the object-attribute detection problem with novel object-attribute pairs. We further design an adaptation approach to leverage CLIP encoders for compositional image classification.
The second part covers a variety of methods built with multimodal transformer models. For image retrieval, we propose a framework that assembles multimodal inputs into sequences with which a multimodal transformer encoder can be fine-tuned. The pre-training of vision-language models (VLMs) is also explored. Specifically, we introduce a fractional intermediate tower that improves the feature expressibility of dual-tower vision-language models. We further design a unified pipeline that allows a VLM to learn from not only vision-language corpora but unimodal visual and linguistic data.
Lastly, we study how to leverage the knowledge of Large Language Models (LLMs) for low-shot image classification, in a data- and computation-efficient way.
Zoom Link: https://usc.zoom.us/j/96814169370?pwd=NkhSYWFKNCsya0lyaUFBVlVDQkI3Zz09Location: Hughes Aircraft Electrical Engineering Center (EEB) - 110
Audiences: Everyone Is Invited
Contact: Zhaoheng Zheng
Event Link: https://usc.zoom.us/j/96814169370?pwd=NkhSYWFKNCsya0lyaUFBVlVDQkI3Zz09
-
PhD Dissertation Defense - Mengxiao Zhang
Mon, Apr 29, 2024 @ 01:30 PM - 03:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Robust and Adaptive Algorithm Design in Online Learning: Regularization, Exploration, and Aggregation
Abstract: In recent years, online learning is becoming a central component in Artificial Intelligence and has been widely applied in many real applications. In this thesis, we focus on designing algorithms for online learning with the two characteristics: robustness and adaptivity. Motivated by the existence of unpredictable corruptions and noises in real-world applications such as E-commerce recommendation systems, robustness is a desired property. It means that the designed algorithm is guaranteed to perform well even in adversarial environments. In contrast, adaptivity complements robustness by enhancing performance in benign environments.In order to achieve robustness and adaptivity, we utilize the following three methodologies, namely regularization, exploration, and aggregation. Regularization method has been widely used in the field of machine learning to control the dynamic of the decisions, which is especially important when facing a possibly adversarial environment. In online learning problems, very often the learner can only observe partial information of the environment, making an appropriate exploration method crucial. Aggregation, a natural idea to achieve adaptivity, combines multiple algorithms that work well in different environments. Though intuitive, this requires non-trivial algorithm design for different online learning problems.In this thesis, we design algorithms for a wide range of online learning problems. We first consider the problem of multi-armed bandits with feedback graphs. Then, we consider more complex problems including linear bandits and convex bandits, which involve an infinite number of actions. We hope that the techniques and algorithms developed in this thesis can help improve the current online learning algorithms for real-world applications. Committee Members:Haipeng Luo (Chair), Vatsal Sharan, Renyuan XuLocation: Ronald Tutor Hall of Engineering (RTH) - 114
Audiences: Everyone Is Invited
Contact: Ellecia Williams
-
PhD Dissertation Defense - Alan Romano
Tue, Apr 30, 2024 @ 09:30 AM - 11:30 AM
Thomas Lord Department of Computer Science
University Calendar
Title: Static Program Analyses for WebAssembly
Committee Members: Weihang Wang (Chair), Chao Wang, and Pierluigi Nuzzo
Date/Time: Tuesday, April 30th, 9:30am - 11:30am
Abstract: WebAssembly is a recent standard for the web that aims to enable high-performance web applications that can run at near-native speeds. The standard has gained attention in both academia and industry for its ability to speed up existing user-facing web applications. Due to its well-defined and sound design, many static program analysis techniques have been developed to accomplish various purposes of WebAssembly analysis. However, we identify gaps in the static program analysis tools of the current WebAssembly ecosystem. We find that current program optimizations applied on WebAssembly modules may lead to diminished performance. We also identify a lack of tools that help developers understand WebAssembly modules through robust binary decompilation. Finally, we find a gap in the ability to analyze cross-language WebAssembly applications across the two languages they are typically implemented in, i.e., WebAssembly and JavaScript.
In this thesis, we present a novel WebAssembly Analysis Framework, or WAF . WAF is a static program analysis framework for WebAssembly modules that consists of multiple intermediate representations. Inspired by frameworks made for Java, the core of our framework lies in our three intermediate representations that each model the WebAssembly module at a different semantic level. This structure enables WAF to serve in multiple use cases, including program optimizations, binary decompilation, cross-language program analysis, and malware detection. We aim to show that our framework can improve static program analysis in the areas that the WebAssembly ecosystem is lacking.Location: Henry Salvatori Computer Science Center (SAL) - 322
Audiences: Everyone Is Invited
Contact: Alan Romano
-
PhD Thesis Proposal - Tian Ye
Tue, Apr 30, 2024 @ 03:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
Title: Enhancing Adversarial Training in Low-Label Regimes
Committee Members: Viktor Prasanna (Chair), Paul Bogdan, Jyotirmoy Deshmukh, Rajgopal Kannan, Cauligi Raghavendra
Data & Time: April 30, 3:00 PM - 4:00 PM Location: EEB 219
Abstract: As machine learning models are increasingly deployed in critical real-world applications, ensuring their robustness against adversarial attacks is essential to prevent potentially severe consequences. Adversarial training, which involves teaching models to recognize and resist adversarial perturbations, is a key strategy for building such robustness. This thesis explores the enhancement of adversarial robustness in scenarios characterized by low-label regimes, where extensive labeled training data are not accessible, by addressing several challenges in existing semi-supervised adversarial training methods. Specifically, the proposed research focuses on: (1) optimizing the generation of adversarial samples to reduce the risk of overfitting, (2) enhancing the reliability of pseudo-labels to mitigate confirmation bias, and (3) simplifying the optimization of training processes to enhance accessibility and efficiency. These improvements will contribute to strengthening the security and functionality of machine learning applications against adversarial threats in a broader range of applications.Location: Hughes Aircraft Electrical Engineering Center (EEB) - 219
Audiences: Everyone Is Invited
Contact: Ellecia Williams
-
PhD Thesis Proposal - Nan Xu
Tue, Apr 30, 2024 @ 04:30 PM - 06:30 PM
Thomas Lord Department of Computer Science
University Calendar
Committee Members: Xuezhe Ma (chair), Muhao Chen, Jonathan May, Ram Nevatia, Daniel O’Leary
Title: Decoding Recipes for Coherent and Factual Text Generation
Abstract: While Large language models (LLMs) have demonstrated increasing power in generating texts, they have also called upon studies on their degeneration problems such as repetition, incoherence, hallucination, etc. My PhD thesis outlines my research aiming to tackle these challenges from the perspective of decoding, which is train-free and driven by models' own understanding of seen and generated texts. Specifically, I focus on 1) reducing undesired repetitions and off-topic generations by analyzing probability distribution of decoding steps for open-ended text generation and 2) mitigating hallucinations by studying models' uncertainty against user prompts for false-premise question answering. Motivated by the emergent ability of Large Vision Language Models (LVLMs) to perceive and understand visual signals, I will also introduce my proposal to mitigate hallucinations with effective decoding strategies given multimodal inputs.
Venue: RTH 306 and Zoom https://usc.zoom.us/j/97468606369?pwd=a2ovTlYweE1neGpTMHFtUlNrcVVnQT09
Date: 04/30/2024, 4:30-6:30PMLocation: Ronald Tutor Hall of Engineering (RTH) - 306
Audiences: Everyone Is Invited
Contact: Ellecia Williams
