Select a calendar:
Filter October Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for October
-
PhD Thesis Proposal - Jun Yan
Mon, Oct 02, 2023 @ 10:15 AM - 11:45 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Jun Yan
Committee Members: Xiang Ren (Chair), Emilio Ferrara, Morteza Dehghani, Robin Jia
Title: Identifying and Mitigating Safety Risks in Language Models
Abstract: In recent years, the rapid advancements in language models have transformed the field of Natural Language Processing, reshaping human technology interactions. As these models become increasingly integrated in various aspects of our daily lives, concerns about their safety risks have also escalated. This thesis proposal outlines my efforts to identify and mitigate safety risks in language models that could lead to system malfunctions and undermine user trust. My research focuses on answering the following key questions 1.How can we expose a models vulnerability by pinpointing robustness challenges? 2.What system failures and harms can adversaries induce through training data poisoning? 3. What strategies can be employed to mitigate risks leading to model malfunction? In conclusion, I will discuss future directions for the development of safer language modelsAudiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/6633659669
-
DREAM Industry Mentorship speaker series
Thu, Oct 05, 2023 @ 12:30 AM - 01:30 PM
USC Viterbi School of Engineering
University Calendar
DREAM (Direct Response to Engineers Aspirations from Mentors) is a mentorship initiative which connects students with high profile industry professionals relevant to students dream pitches, an original leadership communication assignment where students create a vision for their future selves, align their careers around purpose, and build character in the context of growth, reinvention, and constant change. Industry mentors serve as moral exemplars as they discuss how professional challenges present opportunities for character and leadership development. Guest speaker and panels feature professionals from a variety of tech and destination companies.
This event will feature a talk by Alex Kruglov, serial tech entrepreneur and CEO and co-founder of pop.in and former Hulu executive about the evolving media landscape, games, and the human connection.
Location: Ronald Tutor Hall of Engineering (RTH) - 105
Audiences: Everyone Is Invited
Contact: Elisabeth Arnold Weiss
Event Link: https://cglink.me/2nB/r392277
-
PhD Dissertation Defense - Peifeng Wang
Fri, Oct 06, 2023 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Dissertation Defense - Peifeng Wang
Committee Members:Xiang Ren (chair), Muhao Chen, Emilio Ferrara, Dan O Leary
Title: Externalized Reasoning in Language Models for Scalable, Trustworthy AI
Abstract: Language models LMs have traditionally intertwined language skills and reasoning abilities within a single architecture, relying on reckless scaling for effectiveness, but sacrificing interpretability due to their opaque nature. This presentation advocates for externalizing reasoning from the core language proficiency, which enables the development of compact, accessible LMs with enhanced capabilities. The process of externalized reasoning also offers a pathway to explaining model behavior.
I would present three techniques to build LMs with externalized reasoning. First, I introduce an Imagine and Verbalize framework for generative commonsense reasoning, which decomposes a complex generation task into easier subtasks and learns from a diverse set of indirect supervision from multiple domains. Second, I present a knowledge distillation pipeline which prompts a large LM to rationalize for an open domain question and then trains a small LM to reason consistently. Third, I discuss augmenting an LM with working memory for coherent language reasoning by tracking the states of the described world.
Location: Ronald Tutor Hall of Engineering (RTH) - 306
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
-
PhD Thesis Proposal - Fei Wang
Tue, Oct 10, 2023 @ 04:00 PM - 05:30 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Fei Wang
Committee Members: Muhao Chen (Chair), Laurent Itti, Aram Galstyan, Robin Jia, Tianshu Sun
Title: Robust and Context Faithful Language Understanding with (Large) Language Models
Abstract: Large language models (LLMs) have achieved remarkable success in various language understanding tasks. However, their deployment in real world scenarios raises significant accountability concerns. In this talk, I will begin with the contextual faithfulness issue. LLMs often rely on biased parametric knowledge to make unfaithful predictions. I will present a causality driven approach aimed at mitigating entity bias to ensure context faithful NLU. Subsequently, I will introduce the robustness issue against unknown prediction shortcuts. I will demonstrate how to address the issue by proactively mitigating attention biases. Finally, I will outline potential future directions for advancing LLM accountabilityAudiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/92603629078?pwd=cnAyaFNPY1A5QTJ4Ny92K2NJdUlydz09
-
DREAM Industry Mentorship speaker series
Tue, Oct 17, 2023 @ 12:30 PM - 01:30 PM
USC Viterbi School of Engineering
University Calendar
DREAM (Direct Response to Engineers Aspirations from Mentors) connects students with high profile 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 will feature Nick Daze in conversation about his journey from USC student majoring in Theater Arts and English Literature, to tech entrepreneur and CEO at Heirloom, a company building the Future of Self-Sovereign Identity using advanced blockchain technology.
Location: Michelson Center for Convergent Bioscience (MCB) - 101
Audiences: Everyone Is Invited
Contact: Elisabeth Arnold Weiss
Event Link: https://cglink.me/2nB/r392958
-
CS Teaching Faculty Meeting
Mon, Oct 23, 2023 @ 12:00 PM - 02:00 PM
Thomas Lord Department of Computer Science
University Calendar
Meeting for invited full-time Computer Science teaching faculty only. Event details emailed directly to attendees.
Location: Henry Salvatori Computer Science Center (SAL) - 322
Audiences: Invited Faculty Only
Contact: Melissa Ochoa
-
PhD Thesis Proposal - Pei Zhou
Mon, Oct 23, 2023 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Pei Zhou
Committee Members: Xiang Ren (Chair), Jay Pujara, Toby Mintz, Jesse Thomason, Jieyu Zhao
Title: Common Ground Reasoning for Communicative Agents
Abstract: Conversational AIs have received much attention due to the advent of large language model-powered chatbots such as ChatGPT. Recent studies have also focused on the potential of building AI agents that can interact with humans and the world. However, challenges remain unsolved for AI models to become capable communicative agents including understanding implicit intents and reaching goals. This thesis proposal 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 ground knowledge; 2) modeling theory-of-mind capabilities to build goal-driven dialogue agents; and 3) augmenting planning capabilities in multi-round interactions. I will also discuss future directions including can we teach models to self-discover reasoning structures to adapt to unseen scenarios and can models benefit from memory modules to store and generate new insights from prior experienceAudiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/2065614640
-
PhD Thesis Proposal - Xisen Jin
Tue, Oct 24, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Xisen Jin
Committee Members: Xiang Ren (advisor), Jesse Thomason, Robin Jia, Mahdi Soltanolkotabi, Jieyu Zhao
Title: Building Updatable Language Model Systems in the Wild
Abstract: Building updatable language model systems has become a crucial challenge alongside the progress of large language models. In this thesis proposal, I will present my efforts on creating resources, developing efficient methods, and analyzing learning dynamics in updating language models. I will introduce two of my past works, focused on lifelong pretraining of language models and fusing knowledge of multiple models by merging their weights. I will then introduce my on-going study of analyzing and forecasting examples that will be forgotten by model updates to reduce forgetting that happens during the process of model updates.Audiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/92405014194?pwd=RlpqYzNKejZEQ1J1alhQYjBqN3dNZz09
-
DREAM Industry Mentorship speaker series
Tue, Oct 24, 2023 @ 12:30 PM - 01:30 PM
USC Viterbi School of Engineering
University Calendar
DREAM (Direct Response to Engineers Aspirations from Mentors) connects students with high profile 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 will feature global creative, marketing, and tech leader Tom Gilmartin, sharing insights from the evolution of his remarkable career as a creative director working with some of the most valuable brands in the world, including Meta, and his recent co-founding of Verses and Valleys advisory council for non-profits.
Location: Ronald Tutor Hall of Engineering (RTH) - 105
Audiences: Everyone Is Invited
Contact: Elisabeth Arnold Weiss
Event Link: https://cglink.me/2nB/r392960
-
PhD Thesis Proposal - Ang Li
Tue, Oct 24, 2023 @ 02:00 PM - 03:30 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Ang Li
Committee Members: T. K. Satish Kumar (chair), Sven Koenig, Aiichiro Nakano, Emilio Ferrara, and John Carlsson
Title: Revisiting FastMap: New Applications
Abstract: FastMap was first introduced in the Data Mining community for generating Euclidean embeddings of complex objects. In this talk, I will first generalize FastMap to generate Euclidean embeddings of graphs in near-linear time: The pairwise Euclidean distances approximate a desired graph-based distance function on the vertices. I will then apply the graph version of FastMap to efficiently solve various graph-theoretic problems of significant interest in AI: including shortest-path computations, facility location, top-K centrality computations, and community detection and block modeling. I will also present a novel learning framework, called FastMapSVM, by combining FastMap and Support Vector Machines. I will then apply FastMapSVM to predict the satisfiability of Constraint Satisfaction Problems and to classify seismograms in Earthquake ScienceLocation: Hughes Aircraft Electrical Engineering Center (EEB) - 110
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/92891703811?pwd=MmhNQXJCY3ZhMTRlOGp0aWpBZkRsZz09
-
PhD Thesis Defense - John Francis
Wed, Oct 25, 2023 @ 10:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - John Francis
Committee Members: Mike Zyda, Carl Kesselman, Jernej Barbic, Scott Fraser, Kate White
Title: Neural Network Integration of Multiscale and Multimodal Cell Imaging Using Semantic Parts
Abstract: The structural modeling of cells can be accomplished by integrating images of cellular morphology from multiple scales and modalities using a parts based approach. In this thesis, we demonstrate a method for combining the statistical distribution of structures from x-ray tomography and fluorescence microscopy using neural networks to predict the localization of high resolution components in low resolution modalities by using the single cell as a shared unit of transferLocation: Michelson Center for Convergent Bioscience (MCB) - 102
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/5585232420?pwd=QmZSRXI4NkdiMWtXWnp2Q2Q3N1pSdz09
-
PhD Thesis Defense - Georgios Papadimitriou
Wed, Oct 25, 2023 @ 11:00 AM - 01:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Defense - Georgios Papadimitriou
Committee Members: Ewa Deelman (chair), Viktor Prasanna, Aiichiro Nakano
Title: Cyberinfrastructure Management For Dynamic Data Driven Applications
Abstract: Computational science today depends on complex, data intensive applications operating on datasets from a variety of scientific instruments. These datasets may be huge in volume, may have high velocity or both, raising a major challenge of how scientists can analyze these datasets. On the other hand, workflows processing these datasets might need to respond to changes in the processing load e.g, increases in data flow, in order to maintain a steady and predictable turnaround time.
In this thesis we present our efforts to improve the performance of these data intensive application systems. We develop new tools that extend the functionality offered by the CI, and we provide a methodology to capture end to end performance statistics of the data intensive workflows. Additionally, we evaluate how the choices during the acquisition and configuration of resources affect the performance of the data intensive workflows. Finally, we answer the fundamental question of how scientists can manage the CI and apply policies that can help their applications meet their constraints .e.g, turn around time, by avoiding network degradation. We develop methodologies that take place during the planning phase of the workflows, and can reduce their peak network requirements. We also develop active approaches that can be applied and reduce the per workflow network requirements during their execution, using a workflow ensemble manager and application aware software defined flows.
Location: Henry Salvatori Computer Science Center (SAL) - 213
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
-
PhD Thesis Proposal - Jared Coleman
Thu, Oct 26, 2023 @ 12:00 PM - 01:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Jared Coleman
Committee Members: Dr. Bhaskar Krishnamachari (Chair), Dr. Rafael Ferreira da Silva, Dr. Jyotirmoy Deshmukh, Dr. Konstantinos Psounis, Dr. Murali Annavaram
Title: Dispersed Computing in Dynamic Environments
Abstract: Task scheduling 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). In this proposal, we present our existing work to address this gap and identify theoretical and practical research directions that would build upon our previous work to advance the current state-of-the-art in task scheduling for dynamic environmentsLocation: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/92645861253?pwd=NmRaaE5IeXM0b3VHbEpXRUZzT1Yrdz09
-
PhD Thesis Proposal - Avijit Thawani
Tue, Oct 31, 2023 @ 03:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Thesis Proposal - Avijit Thawani
Committee Members: Jay Pujara (advisor), Dani Yogatama, Swabha Swayamdipta, Aiichiro Nakano, Gerard Hoberg
Title: Tokenisation in Language Models: numeracy and beyond
Abstract: The first step for large language models (LLMs) like ChatGPT is to convert text (that humans understand) into indices (that models do). This crucial phase in the Language Modeling pipeline has unfortunately been understudied and is currently achieved by subword segmentation, a manually engineered set of heuristics. We deep dive into case studies where these heuristics fail and our proposed improvements: for example when representing numbers in text, as well as multi-word phrases. Finally, we present an end-to-end tokenized language model that understands both words and numbers better than subwords without any manually engineered heuristic. It also outperforms character-level tokenisation, promising up to 4/6x speed up in inference and training respectivelyLocation: Hughes Aircraft Electrical Engineering Center (EEB) - 110
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
