Select a calendar:
Filter June Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for June
-
PhD Defense - Emily Sheng
Fri, Jun 04, 2021 @ 09:00 AM - 11:00 AM
Thomas Lord Department of Computer Science
University Calendar
PhD Candidate: Emily Sheng
Time: June 4th, 2021, 9AM-11AM
Title: Fairness in Natural Language Generation
Committee: Prof. Prem Natarajan (chair), Prof. Nanyun Peng (chair), Prof. Shri Narayanan, Prof. Yan Liu
Zoom link: https://usc.zoom.us/j/99069448766
Abstract:
Technology for natural language generation (NLG) has advanced rapidly, spurred by advancements in pre-training large models on massive amounts of data and the need for intelligent agents to communicate in a natural manner. While techniques can effectively generate fluent text, they can also produce undesirable societal biases that can have a disproportionately negative impact on already marginalized populations. In this presentation, I emphasize the need for techniques to make language generation applications more fair and inclusive, and further propose a few of these techniques.
The first half of this presentation introduces the problem of societal biases in NLG and how we can use existing and novel quantitative measures as metrics to quantify biases in language generation. I start by introducing a survey and commentary on the existing body of work on fairness in language generation. To meaningfully iterate on techniques that can reduce biases in language generation, I introduce the notion of the regard towards a demographic, use the varying levels of regard towards different demographics as a defining metric for bias in NLG. Through this and other metrics, we can reveal the extent of the biased nature of language generation techniques.
With the analysis and bias quantifiers introduced in the first half, the second half of this presentation focuses on methods to reduce societal biases in NLG techniques. I focus on two methods for controllable generation. The first method builds upon the idea of adversarial triggers to induce societal biases in generated text when input prompts contain mentions of specific demographic groups. The second method is a constrained decoding technique that uses salient n-gram similarity as soft constraints for top-k sampling. We introduce this second method in the context of reducing the disproportionate amount of harmful ad hominem responses faced by marginalized populations in dialogue generation.
WebCast Link: https://usc.zoom.us/j/99069448766
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
Viterbi Wellness Cafe
Thu, Jun 10, 2021 @ 05:00 PM - 05:30 PM
Viterbi School of Engineering Masters Programs
University Calendar
Viterbi Summer Start Graduate Students:
Summer semester is short and midterm can be stressful. Meet with Dr. Yong Park from USC Student Health Center for a relaxing check-in coffee break to help you destress and prepare for the next few weeks!Audiences: Everyone Is Invited
Contact: Juli Legat
-
PhD Defense - Xusen Yin
Mon, Jun 14, 2021 @ 01:30 PM - 03:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Candidate: Xusen Yin
Date: June 14, 2021
Time: 1:30-3pm
Zoom Link:
https://usc.zoom.us/j/98248745757?pwd=VDcvWHpWT0tLSCs0eldsRjhwVGh3UT09
Title:
Generalized Sequential Decision-Making via Language
Committee: Jon May (Chair), Sven Koenig, Shri Narayanan
Abstract:
Many interactive applications, e.g., negotiation, game-playing, personal assistants, and online customer service, require sequential decision-making over natural language communications, a novel but fast-growing domain in Natural Language Processing (NLP).
Unlike other NLP tasks that deal with single sentences or documents---e.g., question answering, machine translation, and sentiment analysis---sequential decision making requires language understanding and inference over sequences of sentences or documents. Moreover, there is usually no direct training objective for these applications compared to typical machine learning tasks. Thus, these tasks require a search in a decision-making space.
Deep Reinforcement Learning (RL) is a common choice for these tasks without explicit or direct targets. It ordinarily needs many iterations to get close to an actual target due to the demand of exploration in the tremendous search space induced by near-infinite natural language responses in dialogue. These explorations usually are composed of many random movements, especially in the initial RL training stage.
However, when placed in an unfamiliar environment, humans know how to solve new problems by applying existing knowledge and skills rather than working randomly. In contrast, computer agents struggle in these new scenarios due to overfitting and lack of common sense.
Can we generalize sequential decision-making agents to novel, even unrelated tasks under the medium of language? We show how to train agents that take beneficial decision sequences from experience and external knowledge for better generalization than standard RL algorithms, using text-based games as a demo environment. We find out that proper dialogue encoding helps the intent understanding, that turning instance knowledge into universal knowledge helps in-domain generalization, that large language models can provide external knowledge rather than learning everything from scratch. Finally, we show that fine-tuned large language models with decision-making ability from one domain can guide RL algorithms towards better exploration and generalization for cross-domain transfer.WebCast Link: https://usc.zoom.us/j/98248745757?pwd=VDcvWHpWT0tLSCs0eldsRjhwVGh3UT09
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
PhD Defense - Sungyong Seo
Fri, Jun 18, 2021 @ 01:00 PM - 02:30 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Candidate: Sungyong Seo
Committee: Prof. Yan Liu (chair), Prof. Xiang Ren, Prof. Antonio Ortega
Date: June 18th, 2021
Time: 1:00-2:30pm
Zoom Link: https://usc.zoom.us/j/7346393285
Title: Physics-aware Graph Networks for Spatiotemporal Physical Systems
Abstract:
While deep neural networks have been successful over a number of applications, it is still challenging to achieve a robust model for physical systems since data-driven learning does not explicitly consider physical knowledge, which should be beneficial for modeling. To leverage domain knowledge for robust learning, I propose various novel methods to incorporate physical knowledge for modeling spatiotemporal observations from physical systems. First, in my talk, I quantify data quality inspired by physical properties of fluids to identify abnormal observations and improve forecasting performance. The second work proposes a regularizer to explicitly impose partial differential equations (PDEs) associated with physical laws to provide an inductive bias in the latent space. The third method focuses on the approximation of spatial derivatives, which are one of the fundamental components of spatiotemporal PDEs. Then, I demonstrate a meta-learning framework to prove that the physics-related quantity is beneficial for fast-adaptation of learnable models on few observations. Finally, I propose spatiotemporal modeling via physics-aware causality, which leverages additional causal information described in PDEs for physical systems. All methods share a common goal: how to integrate physical knowledge with graph networks to model sensor-based physical systems by providing a strong inductive bias.WebCast Link: Link: https://usc.zoom.us/j/7346393285
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
PhD Defense - Nazgol Tavabi
Mon, Jun 21, 2021 @ 10:00 AM - 11:30 AM
Thomas Lord Department of Computer Science
University Calendar
Title: Modeling Dynamic Behaviors in the Wild
PhD Candidate: Nazgol Tavabi
Committee: Kristina Lerman (Chair), Shrikanth (Shri) Narayanan, Bistra Dilkina, Xiang Ren
Thesis abstract:
The abundant real-time data collected from people in the wild creates new opportunities to better understand human behaviors. One example, is temporal data collected from wearable sensors. The ability to analyze this data, offers new opportunities for real-time monitoring of physical and psychological health. Physiological data collected from wearable sensors has been used to detect activities, diagnose illnesses, and analyze habits and personality traits. However, temporal physiological data presents many analytic challenges: the data is multimodal, heterogeneous, noisy; may contain missing values, and long sequences with different lengths. Existing methods for time series analysis and classification are often not suitable for data with these characteristics, nor do they offer interpretability and explainability, a critical requirement in the health domain.
In this thesis, I address some of the challenges in learning representations from these complex temporal data. First, I propose a method based on non-parametric Hidden Markov Models to learn interpretable representations from time series. This method is applied to analyze, cluster, regress and classify multiple datasets. Second, I propose Pattern Discovery with Byte Pair Encoding method to better capture long-term dependencies in lengthy time series, which learns representations by extracting variable length patterns using Byte Pair Encoding compression technique. The proposed model is interpretable, explainable and computationally efficient, and beats state-of-the-art approaches on a real world dataset collected from wearable sensors. Finally, I systematically evaluate how the presence of missing data affects the performance of different state-of-the-art time series classification methods. My work shows how performance of different methods degrades as a function of missing data and, using imputation methods generally does not make a significant difference in the results.
The proposed models and findings, could help better understand and analyze dynamic behaviors within a population and offer new perspectives on monitoring and predicting human behaviors from data collected in the wild.
WebCast Link: https://usc.zoom.us/j/99785051935?pwd=STBuSXpLRnpDUVRQdXJxeGZIZFJZUT09
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
Viterbi New Student Advantage
Tue, Jun 22, 2021 @ 06:00 PM - 07:00 PM
Viterbi School of Engineering Masters Programs
University Calendar
Welcome Viterbi Summer Start Graduate Students! Here is a can't-miss information session for you, mark your calendar and join us on time!
Meet with Kaitlin Harada, Director of Viterbi Connection Center.
Viterbi Career Connections (VCC) offers career-focused support to prepare engineering students for internships, co-ops and full-time employment. We host a variety of technical companies throughout the year with job openings to fill.Audiences: Graduate
Contact: Juli Legat
