Conferences, Lectures, & Seminars
Events for May
-
AI Seminar-Centralities, Communities and Dynamics: The Generalized Laplacian Framework
Fri, May 01, 2015 @ 11:00 AM - 12:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: Xiaoran Yan , USC/ISI
Talk Title: Centralities, Communities and Dynamics: The Generalized Laplacian Framework
Series: Artificial Intelligence Seminar
Abstract: The interplay between a dynamic process and topology of a network on which it unfolds affects observed network structure. In this talk, we examine the impact of this the interaction on the identification of central nodes and communities in networks. We introduce the generalized Laplacian framework which extends the traditional Laplacian beyond simple diffusion. By mathematically relating dynamic processes to random walks on a transformed network, the generalized Laplacian formulation unifies different measures of centrality and community quality. Thus, a node's centrality describes its participation in the dynamics taking place on the network, and communities are groups of nodes that interact more frequently with each other according to the rules of the dynamic process. We prove that the classic Cheeger's inequality, which relates the spectrum of the Laplacian matrix to the conductance of the best cluster in the network, can be extended to this generalized setting, providing a method for fast graph partitioning under any dynamics. We demonstrate empirically that different dynamic processes lead to divergent views of structure of synthetic and real-world networks.
Biography: Xiaoran Yan is a Postdoctoral Research Associate at Kristina Lerman's group at ISI-USC. His work focuses on modelling and analyzing dynamics on networks as well as the underlying topological structures. He received a Ph.D. in Computer Science at University of New Mexico.
Host: Kristina Lerman
Webcast: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=9e22bdf6c8be40a3a5f7abdb6c3fef251dLocation: Information Science Institute (ISI) - 11th Flr. Conf Rm # 1135, Marina Del Rey
WebCast Link: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=9e22bdf6c8be40a3a5f7abdb6c3fef251d
Audiences: Everyone Is Invited
Contact: Peter Zamar
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor. -
Learning Longer Memory in Recurrent Networks
Tue, May 12, 2015 @ 11:00 AM - 12:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: Tomas Mikolov , Research Scientist at Facebook AI Reseach
Talk Title: Learning Longer Memory in Recurrent Networks
Series: AISeminar
Abstract: Recurrent neural network is a powerful model that learns temporal patterns in sequential data. For a long time, it was believed that recurrent networks are difficult to train using simple optimizers, such as stochastic gradient descent, due to the so-called vanishing gradient problem. In this talk, I will show that learning longer term patterns in real data, such as in natural language, is perfectly possible using gradient descent. This is achieved by using a slight structural modification of the simple recurrent neural network architecture. Some of the hidden units are encouraged to change their state slowly by constraining part of the recurrent weight matrix to be close to identity, thus forming kind of a longer term memory. We evaluate our model in language modeling experiments, where we obtain similar performance to the much more complex Long Short Term Memory (LSTM) networks. This is a joint work with Armand Joulin, Sumit Chopra, Michael Mathieu and Marc'Aurelio Ranzato.
Biography: Tomas Mikolov is a research scientist at Facebook AI Research. His work includes introduction of recurrent neural networks to statistical language modeling (published as open-source RNNLM toolkit), and an efficient algorithm for estimating word representations in continuous space (the Word2vec project). His current interest is in developing techniques and datasets that would help to advance research towards artificial intelligence systems capable of natural communication with people.
Website: https://research.facebook.com/researchers/643234929129233/tomas-mikolov/
Host: Ashish Vaswani
Webcast: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=19140806ae5e4116ab2644b1c1d86bbe1dLocation: Information Science Institute (ISI) - 1135
WebCast Link: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=19140806ae5e4116ab2644b1c1d86bbe1d
Audiences: Everyone Is Invited
Contact: Alma Nava / Information Sciences Institute
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor. -
AI Seminar: Computerized Search for Causal Relations in High Dimensional Data: Some Results and Many Problems
Wed, May 13, 2015 @ 11:00 AM - 12:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: Clark Glymour , Alumni University Professor, Carnegie Mellon University
Talk Title: Computerized Search for Causal Relations in High Dimensional Data: Some Results and Many Problems
Series: Artificial Intelligence Seminar
Abstract: I will briefly review the graphical causal model framework and describe some of the search strategies that have proved practical in small dimensional problems. Then I will describe some of the modifications we have recently pursued at Carnegie Mellon to allow search in high dimensional problems, e.g, 50.000 - 1,000,000 variables, with sample sizes orders of magnitude smaller, and some of the many problems we have not satisfactorily solved.
Biography: Clark Glymour is the Alumni University Professor in the Department of Philosophy at Carnegie Mellon University. He is also a senior research scientist at the Florida Institute for Human and Machine Cognition. He is the founder of the Philosophy Department at Carnegie Mellon University, a Guggenheim Fellow, a Fellow of the Center for Advanced Study in Behavioral Sciences a Phi Beta Kappa lecturer, and is a Fellow of the statistics section of the AAAS. Glymour and his collaborators created the causal interpretation of Bayes nets and developed an automated causal inference algorithm implemented as software named TETRAD. His areas of interest include epistemology (particularly Android epistemology), machine learning, automated reasoning, psychology of judgment, and mathematical psychology.
Host: Kun Zhang
More Info: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=ebaa5ed5e1444cbfa7aea272f321509d1d
Webcast: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=ebaa5ed5e1444cbfa7aea272f321509d1dLocation: Information Science Institute (ISI) - 11th Flr Conf Rm # 1135, Marina Del Rey
WebCast Link: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=ebaa5ed5e1444cbfa7aea272f321509d1d
Audiences: Everyone Is Invited
Contact: Peter Zamar
Event Link: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=ebaa5ed5e1444cbfa7aea272f321509d1d
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor. -
NL Seminar- Exploring LDA: Parallel Inference and Model Selection
Fri, May 15, 2015 @ 03:00 PM - 04:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: Dehua Cheng, (USC/Melady)
Talk Title: Exploring LDA: Parallel Inference and Model Selection
Series: Natural Language Seminar
Abstract: Latent Dirichlet allocation (LDA) and its Bayesian nonparametric generalization hierarchical Dirichlet processes (HDP) have been proven successful in modeling large, complex, real-world domains. However, inference on LDA/HDP is challenging and it has received notable attention from the researchers. In this talk, we present two algorithmic advances for LDA/HDP inference by examining their mathematical properties. We will first present an effective parallel Gibbs sampling algorithm for LDA/HDP by exploring the equivalency between the Dirichlet-multinomial hierarchy and the Gamma-Poisson hierarchy. Secondly, we will show how to provably select the number of topics for LDA by studying the spectral space of its second order moments (bi-gram statistics).
Biography: Dehua Cheng is a third year Ph.D. student in the CS department at USC, advised by Professor Yan Liu. Prior to that, he received his B.S. degree in Mathematics and Physics from Tsinghua University, China. His research interests include randomized numerical algorithm in machine learning and parallel inference for probabilistic graphical model.
Host: Nima Pourdamghani and Kevin Knight
More Info: http://nlg.isi.edu/nl-seminar/
Location: Information Science Institute (ISI) - 6th Flr Conf Rm # 689, Marina Del Rey
Audiences: Everyone Is Invited
Contact: Peter Zamar
Event Link: http://nlg.isi.edu/nl-seminar/
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor. -
Artificial vs AuthenticThe Art of Language Invention
Tue, May 19, 2015 @ 03:00 PM - 04:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: David Peterson , HBOs Game of Thrones
Talk Title: Artificial vs Authentic:The Art of Language Invention
Series: Artificial Intelligence Seminar
Abstract: As long as humans have been using language, humans have been inventing language. Linguistic creativity has taken many different forms over the years, but it's made its greatest strides in the past 25 years. For languages like Esperanto and Volapük, functionality and unambiguity were lofty design goals. Today, with languages created for entirely different purposes, functionality is considered a prerequisite, and unambiguity a major design flaw. In this talk, David Peterson discusses naturalistic language creation, and the emerging art form known as conlanging.
Biography: David Peterson is a language creator and author. Since 2009, he's been working on HBO's Game of Thrones, having created the Dothraki and Valyrian languages. Since then, he's gone on to work on a number of other projects, including Syfy's Defiance, Syfy's Dominion, Marvel's Thor: The Dark World, the CW's Star-Crossed, the CW's The 100, and Showtime's Penny Dreadful. He authored the book Living Language Dothraki, an introductory guide to the Dothraki language, and in September, he'll be publishing The Art of Language Invention with Penguin Random House.
Host: Ashish Vaswani
More Info: http://webcasterms1.isi.edu/mediasite/SilverlightPlayer/Default.aspx?peid=714b19be3e114ca79ddb3ccfb55366f01d
Webcast: http://webcasterms1.isi.edu/mediasite/SilverlightPlayer/Default.aspx?peid=714b19be3e114ca79ddb3ccfb55366f01dLocation: Information Science Institute (ISI) - 11th Flr Conf Rm s1135 & 1137 Marina Del Rey
WebCast Link: http://webcasterms1.isi.edu/mediasite/SilverlightPlayer/Default.aspx?peid=714b19be3e114ca79ddb3ccfb55366f01d
Audiences: Everyone Is Invited
Contact: Peter Zamar
Event Link: http://webcasterms1.isi.edu/mediasite/SilverlightPlayer/Default.aspx?peid=714b19be3e114ca79ddb3ccfb55366f01d
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor. -
NL Seminar- How to Memorize a Random 60-Bit String
Fri, May 22, 2015 @ 03:00 PM - 04:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: Marjan Ghazvininejad, USC/ISI
Talk Title: How to Memorize a Random 60-Bit String
Series: Natural Language Seminar
Abstract: User-generated passwords tend to be memorable, but not secure. A random, computer-generated 60-bit string is much more secure. However, users cannot memorize random 60-bit strings. In this paper, we investigate methods for converting arbitrary bit strings into English word sequences (both prose and poetry), and we study their memorability and other properties.
Biography: Marjan Ghazvininejad is a second year PhD student in Computer Science at University of Southern California (USC). She is working with Professor Kevin Knight at the Information Sciences Institute (ISI). She is interested in natural language processing, especially the application of machine learning techniques in this area.
Host: Nima Pourdamghani and Kevin Knight
More Info: http://nlg.isi.edu/nl-seminar/
Location: Information Science Institute (ISI) - 6th Flr Conf Rm # 689, Marina Del Rey
Audiences: Everyone Is Invited
Contact: Peter Zamar
Event Link: http://nlg.isi.edu/nl-seminar/
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor. -
NL Seminar-How to Make a Frenemy: Multitape FSTs for Portmanteau Generation
Fri, May 29, 2015 @ 03:00 PM - 04:00 PM
Information Sciences Institute
Conferences, Lectures, & Seminars
Speaker: Aliya Deri, USC/ISI
Talk Title: How to Make a Frenemy: Multitape FSTs for Portmanteau Generation
Abstract: A portmanteau is a type of compound word that fuses the sounds and meanings of two component words; for example, "frenemy" (friend + enemy) or smog (smoke + fog). We develop a system, including a novel multitape FST, that takes an input of two words and outputs possible portmanteaux. Our system is trained on a list of known portmanteaux and their component words, and achieves 45% exact matches in cross-validated experiments.
Biography: Aliya Deri is a PhD candidate at USC/ISI.
Host: Nima Pourdamghani and Kevin Knight
More Info: http://nlg.isi.edu/nl-seminar/
Location: 6th Flr Conf Rm # 689, Marina Del Rey
Audiences: Everyone Is Invited
Contact: Peter Zamar
Event Link: http://nlg.isi.edu/nl-seminar/
This event is open to all eligible individuals. USC Viterbi operates all of its activities consistent with the University's Notice of Non-Discrimination. Eligibility is not determined based on race, sex, ethnicity, sexual orientation, or any other prohibited factor.
