All Seminars
| Title: Computational Large-Scale Continuous Optimization, Uncertainty and Robustness |
|---|
| Colloquium: N/A |
| Speaker: Somayeh Moazeni of Princeton University |
| Contact: James Nagy, nagy@mathcs.emory.edu |
| Date: 2014-02-10 at 4:00PM |
| Venue: MSC W303 |
| Download Flyer |
| Abstract: Optimal decisions often rely on assumptions about the models and their associated parameter values. Therefore, it is essential to assess the suitability of these assumptions and to understand the sensitivity of outcomes when they are altered. More importantly, appropriate approaches should be developed to achieve a robust solution. In this talk, we first present a sensitivity analysis on parameter values as well as model specification of a problem in portfolio management, namely the optimal portfolio execution problem. We then propose more robust solution techniques and models including regularized robust optimization for convex optimization programs and computational stochastic optimization. Extensions of these approaches for energy storage operational management and electricity price modeling are discussed. |
| Title: Independent Sets in Hypergraphs |
|---|
| Colloquium: N/A |
| Speaker: Dhruv Mubayi of The University of Illinois at Chicago |
| Contact: Dwight Duffus, dwight@mathcs.emory.edu |
| Date: 2014-02-07 at 4:00PM |
| Venue: MSC W201 |
| Download Flyer |
| Abstract: Abstract: The problem of determining the independence number of (hyper)graphs has tight connections to questions in discrete geometry, coding theory, number theory, theoretical computer science and combinatorics. One of the most famous early examples is the result of Komlos-Pintz-Szemeredi from 1982 on the independence number of 3-uniform hypergraphs which made important progress on the decades old Heilbronn problem. I will begin by explaining this result and some of these connections. I will then describe recent work in this area which shows that hypergraphs have a significantly different behavior than graphs when it comes to independent sets. This answers a question posed by Ajtai-Erdos-Komlos-Szemeredi (1981), and disproves conjectures of deCaen (1986), Frieze and the speaker (2007), and several others. |
| Title: Continuous analogues of methods used to calculate component groups of Jacobians |
|---|
| Seminar: Algebra |
| Speaker: Joe Rabinoff of Georgia Tech |
| Contact: David Zureick-Brown, dzb@mathcs.emory.edu |
| Date: 2014-02-04 at 4:00PM |
| Venue: W302 |
| Download Flyer |
| Abstract: Let K be complete, discretely-valued field and let X be a smooth projective K-curve equipped with a semistable model over the valuation ring. A series of classical theorems, mostly due to Raynaud, give two ways of calculating the component group of the Jacobian J of X: one using the intersection matrix on the special fiber of the model of X, and the other using cycles on its incidence graph G. These calculations can be interpreted in terms of divisors on G (in the sense of Baker-Norine) and the uniformization theory of G, respectively. If K is complete and non-Archimedean but not discretely valued, these theorems are no longer applicable, as Néron models do not exist in this situation. Replacing the component group with the skeleton of J (in the sense of Berkovich), a principally polarized real torus canonically associated to J, and the incidence graph with a skeleton Gamma of X, a metric graph, we will prove "continuous" analogues of these theorems. Specifically, we will show that the Jacobian of Gamma is canonically identified with the skeleton of J as principally polarized real tori, in a way that is compatible with the descriptions of the two Jacobians in terms of divisors and in terms of uniformizations. As a consequence, we will show that, when K is algebraically closed, essentially any piecewise-linear function on Gamma is the restriction to Gamma of $-\log |f|$, where f is a nonzero rational function on X. |
| Title: Numerical Methods for Hyperelastic Image Registration |
|---|
| Colloquium: N/A |
| Speaker: Lars Ruthotto of University of British Columbia |
| Contact: James Nagy, nagy@mathcs.emory.edu |
| Date: 2014-01-31 at 3:00PM |
| Venue: MSC W201 |
| Download Flyer |
| Abstract: Image registration is an essential task in almost all areas involving imaging techniques. The goal of image registration is to find geometrical correspondences between two or more images. Image registration is commonly phrased as a variational problem that is known to be ill-posed and thus regularization is commonly used to ensure existence of solutions and/or introduce prior knowledge about the application in mind. This talk presents a nonlinear regularization functional based on the theory of hyperelastic materials, which overcomes limitations of the most commonly used linear elastic model. In particular, the hyperelastic regularization functional guarantees that solutions to the variational problem exist and are one-to-one correspondences between the images, which is a key concern in most applications. The focus of this talk is on accurate and fast numerical methods for solving hyperelastic image registration problems. Further, the potential of hyperelastic schemes is demonstrated for real-life medical imaging problems. |
| Title: Randomized Block Coordinate Gradient Methods for a Class of Structured Nonlinear Programming |
|---|
| Colloquium: N/A |
| Speaker: Zhaosong Lu of Simon Fraser University |
| Contact: James Nagy, nagy@mathcs.emory.edu |
| Date: 2014-01-29 at 3:00PM |
| Venue: MSC W201 |
| Download Flyer |
| Abstract: Nowadays a class of huge-scale structured optimization problems arise in some emerging areas such as machine learning. They can be reformulated as minimizing the sum of a smooth and block separable nonsmooth functions. For these problems, it is prohibitive to evaluate the full gradient of the smooth component of the objective function due to huge dimensionality and hence the usual gradient methods cannot be efficiently applied. Nevertheless, its partial gradients can often be computed much more cheaply. In this talk we study a randomized block coordinate gradient (RBCG) method for solving this class of problems. At each iteration this method randomly picks a block, and solves a proximal gradient subproblem over the subspace defined by the block that only uses a partial gradient and usually has a closed-form solution. We present new iteration complexity results for this method when applied to convex problems. We also propose a nonmonotone RBCG method for solving a class of nonconvex problems with the above structure, and establish their global convergence and iteration complexity. In addition, we present new complexity results for the accelerated RBCG method proposed by Nesterov for solving unconstrained convex optimization problems. Finally, we discuss the application of these methods for solving some support vector machine problems and report some computational results. (This is a joint work with Lin Xiao at Microsoft Research Redmond.) |
| Title: Something special... |
|---|
| Seminar: Algebra |
| Speaker: Ken Ono of Emory |
| Contact: David Zureick-Brown, dzb@mathcs.emory.edu |
| Date: 2014-01-28 at 3:00PM |
| Venue: W304 |
| Download Flyer |
| Abstract: |
| Title: More examples of non-rational adjoint groups |
|---|
| Seminar: Algebra |
| Speaker: Nivedita Bhaskhar of Emory University |
| Contact: David Zureick-Brown, dzb@mathcs.emory.edu |
| Date: 2014-01-21 at 4:00PM |
| Venue: W302 |
| Download Flyer |
| Abstract: A k-variety is said to be rational if its function field is purely transcendental over k. The first example of a non-rational adjoint k-group PSO(q) was given by Merkurjev as a consequence of his computations of R-equivalence classes of adjoint classical groups. The quadratic form in question has non-trivial discriminant which property is used crucially in the proof. Gille provided the first example of a quadratic form of trivial discriminant whose associated adjoint group is non-rational. In this talk we give a recursive construction to produce examples of $k_n$-quadratic forms $q_n$ in the n-th power of the fundamental ideal in the Witt ring whose corresponding adjoint groups PSO($q_n$) are not (stably) rational. |
| Title: Hierarchy and Structure: Nonparametric models for space, language, and relations |
|---|
| Seminar: Computer Science |
| Speaker: Alex Smola of Google/CMU |
| Contact: Eugene Agichtein, eugene@emory.edu |
| Date: 2014-01-09 at 3:00PM |
| Venue: MSC W301 |
| Download Flyer |
| Abstract: Latent variable models are a powerful tool for analyzing structured data. They are well suited to capture documents, location and preference information. That said, often a simple hierarchical model is insufficient for modeling observations since real data tends to be more nuanced in some aspects rather than others. In other words, descriptions work best if they allow for variable depth and refined descriptions. Models such as the nested Chinese Restaurant Franchise address these issues. I will present examples of their application to location inference for Twitter and structured recommender systems. |
| Title: Computational Methods for Centrality Measurements in Complex Networks |
|---|
| Defense: Dissertation |
| Speaker: Christine Klymko of Emory University |
| Contact: Christine Klymko, cklymko@emory.edu |
| Date: 2013-12-10 at 3:30PM |
| Venue: W302 |
| Download Flyer |
| Abstract: |
| Title: Scalable and Privacy-Preserving Searchable Cloud Data Services |
|---|
| Seminar: Computer Science |
| Speaker: Ming Li of Utah State University |
| Contact: Li Xiong, lxiong@mathcs.emory.edu |
| Date: 2013-12-09 at 4:00PM |
| Venue: MSC W301 |
| Download Flyer |
| Abstract: Cloud computing is envisioned as the next generation architecture of IT enterprises, which provides convenient remote access to massively scalable data storage and application services. Despite the cloud’s promise for huge potential economical savings, its benefits may not be fully realized, due to wide public concerns that users’ private data may be involuntarily exposed to or mishandled by the cloud providers. Although end-to-end encryption has been proposed as a promising solution for secure cloud data storage, how to effectively support flexible data utilization such as searches over encrypted cloud data becomes a primary challenge, which is the key toward building full-fledged privacy-assured cloud data storage. In this talk, I will first identify the system requirements and challenges in privacy-preserving searchable outsourced cloud data services, that is to simultaneously achieve privacy assurance (data and query confidentiality), practical efficiency (scalable with large volumes of data), and high usability (flexible query functionalities). Among these goals, privacy and the other two are often in conflict with each other and our research aims at finding a better tradeoff. As an example, I will present our recent work on privacy-preserving multi-keyword ranked search supporting similarity-based ranking. The proposed approach integrates novel cryptographic primitives with information-retrieval principles and efficient data structures. A “best-effort” privacy model is adopted while much faster-than-linear search time is achieved in an empirical sense. Finally, I will outline some future challenges that need to be resolved to make privacy-preserving searchable cloud data service a reality.\\ \\ Bio:\\ Ming Li is an Assistant Professor in the Computer Science Department at Utah State University. He received his Ph.D. in Electrical and Computer Engineering from Worcester Polytechnic Institute in 2011. His current main research interest is cyber security and privacy, with emphases on security and privacy in cloud computing and big data, security in wireless networks and cyber-physical systems. |