CMS Upcoming SeminarsCMS Upcoming Seminar Feed
http://www.cms.caltech.edu/seminars.rss
enCMS-EE Partners Tech Talk: TBDclairer@caltech.edu (Claire Ralph)CMS-EE Partners Tech Talk<strong>Speaker(s):</strong> <br><strong>Location:</strong> Annenberg 105<br><p></p>Mon, 24 Feb 2020 12:00:00 -0800http://www.cms.caltech.edu/events/87004Rigorous Systems Research Group (RSRG) Seminar: Distributed decision making in networked systems: from optimization to reinforcement learningsuyu@caltech.edu (Yu Su)Rigorous Systems Research Group (RSRG) Seminar<strong>Speaker(s):</strong> Guannan Qu (Caltech)<br><strong>Location:</strong> Annenberg 121<br><p>Cyber-physical systems such as the power grid, Internet of Things (IoT), and transportation systems are increasingly adding large numbers of devices with sensing capabilities. This results in an explosion of data, and calls for a rethinking of traditional control/optimization theory, especially how to incorporate learning techniques to use the data for control/optimization purposes, as well as how to cope with the challenges caused by the sheer scale of the network. In the first part of the talk, we focus on distributed optimization, where the nodes seek to minimize a global loss function, which is a sum of their local loss functions formed by local data sets. This is a classical setting and our results provide the fastest known gradient-based distributed algorithm. In the second part, we go beyond static optimization and investigate how data can be directly used to design control policy. Particularly, we study reinforcement learning (RL) for localized control of networked systems. Despite its wide-ranging successes, the application of RL in the multi-agent systems has proven to be challenging due to scalability issues. Harnessing the network structure, we develop a Scalable Actor-Critic framework to learn an optimal local policy in a scalable manner. This result represents the first approach that provably addresses the issue of scalability in the context of multi-agent RL.<br></p>Mon, 24 Feb 2020 14:00:00 -0800http://www.cms.caltech.edu/events/88035IQI Weekly Seminar: Quantum East model: localization, non-thermal eigenstates and slow dynamicsbjleung@caltech.edu (Bonnie Leung)IQI Weekly Seminar<strong>Speaker(s):</strong> Nicola Pancotti (Max Planck)<br><strong>Location:</strong> Annenberg 121<br><p><b>Abstract</b>: We study in detail the properties of the quantum East model, an interacting quantum spin chain inspired by simple kinetically constrained models of classical glasses.</p><p>Through a combination of analytics, exact diagonalization and tensor network methods we show the existence of a fast-to-slow transition throughout the spectrum that follows from a localization transition in the ground state.</p><p>On the slow side, we explicitly construct a large (exponential in size) number of non-thermal states which become exact finite-energy-density eigenstates in the large size limit,</p><p>as expected for a true phase transition.</p><p>A ``super-spin'' generalization allows us to find a further large class of area-law states proved to display very slow relaxation.</p><p>Under slow conditions, many eigenstates have a large overlap with product states and can be approximated well by matrix product states at arbitrary energy densities.</p><p>We discuss implications of our results for slow thermalization and non-ergodicity more generally for quantum East-type Hamiltonians and their extension in two or higher dimensions.</p><p></p><p></p><p></p><p></p><p></p><p></p>Tue, 25 Feb 2020 15:00:00 -0800http://www.cms.caltech.edu/events/88020TCS+ Talk: MIP*=REbjleung@caltech.edu (Bonnie Leung)TCS+ Talk<strong>Speaker(s):</strong> Henry Yuen (University of Toronto)<br><strong>Location:</strong> Annenberg 322<br><p><b>Abstract:</b> MIP* denotes the class of problems that admit interactive proofs with quantum entangled provers. It has been an outstanding question to characterize the complexity of MIP*. Most notably, there was no known computable upper bound on this class.We show that MIP* is equal to the class RE, the set of recursively enumerable languages. In particular, this shows that MIP* contains uncomputable problems. Through a series of known connections, this also yields a negative answer to Connes' Embedding Problem from the theory of operator algebras. In this talk, I will explain the connection between Connes' Embedding Problem, quantum information theory, and complexity theory. I will then give an overview of our approach, which involves reducing the Halting Problem to the problem of approximating the entangled value of nonlocal games.<br>Joint work with Zhengfeng Ji, Anand Natarajan, Thomas Vidick, and John Wright.</p>Wed, 26 Feb 2020 10:00:00 -0800http://www.cms.caltech.edu/events/87703CMX Lunch Seminar: An Optimal Transport Perspective on Uncertainty Propagationjbrink@caltech.edu (Jolene Brink)CMX Lunch Seminar<strong>Speaker(s):</strong> Amir Sagiv (Columbia University)<br><strong>Location:</strong> Annenberg 213<br><p> In many scientific areas, a deterministic model (e.g., a differential equation) is equipped with parameters. In practice, these parameters might be uncertain or noisy, and so an honest model should account for these uncertainties and provide a statistical description of the quantity of interest. Underlying this computational problem is a fundamental question - If two "similar" functions push-forward the same measure, are the new resulting measures close, and if so, in what sense? In this talk, I will first show how the probability density function (PDF) can be approximated, and present applications to nonlinear optics. We will then discuss the limitations of PDF approximation, and present an alternative Wasserstein-distance formulation of this problem, which through optimal-transport theory yields a simpler theory. </p>Wed, 26 Feb 2020 12:00:00 -0800http://www.cms.caltech.edu/events/86832Center for Social Information Sciences (CSIS) Seminar: Multiple Imputation for Large Multiscale Data with Linear Constraintsmmartin@caltech.edu (Mary Martin)Center for Social Information Sciences (CSIS) Seminar<strong>Speaker(s):</strong> Jian Cao (Caltech)<br><strong>Location:</strong> Baxter B125<br><p>Abstract: We present a new method that is capable of handling both missing and suppressed value problems for large multiscale data sets, such as the Quarterly Census of Employment and Wages (QCEW) from the U.S. Bureau of Labor Statistics. Existing multiple imputation methods are hard to scale for such data sets. This particularly acute in the case of QCEW, with as many as 1.5 billion observations aggregated along three different scales (industry structure, geographic levels, and time). Our method incorporates three innovations. First, we improve the accuracy of the Bootstrapping-based Expectation Maximization method (King et al. 2010), a state-of-the-art multiple imputation method, by utilizing the extra information from the singular covariance matrix and taking into account of the multiscale data structure. Second, we introduce a quasi-Monte Carlo technique to accelerate convergence. Third, we develop a parallel sequential approach that partitions the large data set into quasi-independent small data sets according to the data structure and patterns of suppressed and missing observations. We demonstrate that our new method improves speed and accuracy. Moreover, it can be applied to large data sets with complicated multiscale structures.</p>Fri, 28 Feb 2020 12:00:00 -0800http://www.cms.caltech.edu/events/87078IQIM Postdoctoral and Graduate Student Seminar: Lightning Round - APS style talksmarciab@caltech.edu (Marcia Brown)IQIM Postdoctoral and Graduate Student Seminar<strong>Speaker(s):</strong> ()<br><strong>Location:</strong> East Bridge 114<br><p><b>February 28th</b> we will host a lightning round seminar, with <b>three</b> <b>10+2 minute talks</b> (10 minutes of talk and 2 minutes of audience questions). We will have a moderator keeping time (and cutting you off promptly at 10 min - just like in your real talk at APS!).</p><p><b>If you are interested in giving one of the talks,</b> please contact the IQIM seminar committee by emailing (<a href="mailto:iqim_seminar_committee@caltech.edu">iqim_seminar_committee@caltech.edu</a>) by<b> Monday 24th</b>. Let us know the title of your talk and abstract, and also if you've ever presented at a conference before. Rookies are welcome and encouraged to apply!</p>Fri, 28 Feb 2020 12:00:00 -0800http://www.cms.caltech.edu/events/88048CMS-EE Partners Tech Talk: TBDclairer@caltech.edu (Claire Ralph)CMS-EE Partners Tech Talk<strong>Speaker(s):</strong> <br><strong>Location:</strong> Annenberg 105<br><p></p>Mon, 02 Mar 2020 12:00:00 -0800http://www.cms.caltech.edu/events/87005H.B. Keller Colloquium: Learning Methods for Decentralized Decision Making in IoT/CPS Systemsdbohler@caltech.edu (Diana Bohler)H.B. Keller Colloquium<strong>Speaker(s):</strong> Urbashi Mitra (University of Southern California)<br><strong>Location:</strong> Annenberg 105<br><p></p><p>The modern instantiation of a sensor network is a cyberphysical (CPS) system where CPS subsystems can be interconnected by a shared communication network of limited bandwidth. A common problem in CPS networks is the sensing and communication of spatio-temporal signals. However, there are fundamental differences between estimation (sensing) and communication. For example, the type of signal one would design to optimize sensing is very different from that for optimized communication. In this talk, we explore some of these differences and discuss how joint communication and sensing should occur in different problem settings. In particular, we examine problems where multiple sensors make observations and must share the communication medium to transmit these signals to a fusion center that will endeavor to perform remote estimation all of the sensed signals. A new class of remote estimation problems, where the communication resources are allocated dynamically based on the observations at the sensors, rather than purely on their statistical description is examined. We address the optimal design of a collision avoidance policy by selecting the most informative sensor to transmit at a time. First, we will establish person-by-person optimal policies for the scheduling of sensors making Gaussian observations. Then, we will show how our theoretical results can be applied to design scheduling policies where the joint probability density of the observations is unknown. We will extend our results to the case when the scheduler uses an energy harvesting battery as well as to the case of purely decentralized decision making. Time permitting, extensions to the case of microbial decision making in quorum sensing networks will be discussed.</p>Mon, 02 Mar 2020 16:00:00 -0800http://www.cms.caltech.edu/events/87999Special CMX Seminar: Mathematical and Computational Aspects of Imaging with Wavesjbrink@caltech.edu (Jolene Brink)Special CMX Seminar<strong>Speaker(s):</strong> Liliana Borcea (University of Michigan)<br><strong>Location:</strong> Annenberg 213<br><p> </p><p>Wave based imaging is an inverse problem for a wave equation or a system of </p><p>equations with a wide range of applications in nondestructive testing of structures such as airplane wings, ultrasound for medical diagnosis, radar, sonar, geophysical exploration, etc. It seeks to determine scattering structures in a medium, modeled mathematically by a reflectivity function, from data collected by sensors that probe the medium with signals and measure the resulting waves. Most imaging methods formulate the inverse problem as a least squares data fit optimization, and assume a linear mapping between the unknown reflectivity and the data. The linearization, known as the Born (single scattering) approximation is not accurate in strongly scattering media, so the reconstruction of the reflectivity may be poor. I will describe a new inversion methodology that is based on a reduced order model approach. This borrows ideas from dynamical systems, where the reduced order model is a projection of an operator, called wave propagator, which describes the propagation of the waves in the unknown medium. I will explain how such a reduced order model can be constructed from measurements at the sensors and then I will show how it can be used for improving the existing inversion methodology.</p>Tue, 03 Mar 2020 16:30:00 -0800http://www.cms.caltech.edu/events/87511EE Systems Seminar: Empirical Risk Minimization in High-dimensions: Asymptotics, Optimality and Double Descentlchavarr@caltech.edu (Liliana Chavarria)EE Systems Seminar<strong>Speaker(s):</strong> Christos Thrampoulidis (UC Santa Barbara)<br><strong>Location:</strong> Moore B280<br><p><b>ABSTRACT</b> At the heart of contemporary statistical signal processing problems, as well as modern machine-learning practices, lie high-dimensional inference tasks in which the number of unknown parameters is of the same order as (and often larger than) the number of observations.</p><p>In this talk, I describe a framework based on Gaussian-process inequalities to <i>sharply</i> characterize the statistical performance of convex empirical risk minimization in high dimensions. By focusing on the simple, yet highly versatile, model of binary linear classification, I will demonstrate that, albeit challenging, sharp results are advantageous over loose order-wise bounds. For instance, they lead to precise answers to the following questions: When are training data linearly separable? Is least-squares bad for binary classification? What is the best convex loss function? Is double descent observed in linear models and how do its features (such as the transition threshold and global minima) depend on the training data and on the learning procedure?</p><p>Many of the ideas and technical tools of our work originate from the study of sharp phase-transitions in compressed sensing. Throughout the talk, I will highlight how our results relate to and advance this literature.</p><p><b>BIO</b> Christos Thrampoulidis is an Assistant Professor in the ECE Department at UC Santa Barbara since July 2018. His research interests include statistical signal processing, optimization, and, machine learning. Before joining UCSB, Dr. Thrampoulidis was a Postdoctoral Researcher at the Research Laboratory of Electronics at MIT. He received his M.Sc. and Ph.D. degrees in Electrical Engineering from Caltech in 2012 and 2016, respectively, and his Diploma of electrical and computer engineering from the University of Patras in Greece in 2011. He is a recipient of the 2014 Qualcomm Innovation Fellowship.</p>Wed, 04 Mar 2020 16:00:00 -0800http://www.cms.caltech.edu/events/88033Finance Seminar: Top Wealth in the United States: New Estimates and Implications for Taxing the Richsabrina@hss.caltech.edu (Sabrina Hameister)Finance Seminar<strong>Speaker(s):</strong> Eric Zwick (University of Chicago)<br><strong>Location:</strong> Baxter B125<br><p>Abstract: This paper uses administrative tax data to estimate top wealth in the United States. We build on the capitalization approach in Saez and Zucman (2016) while accounting for heterogeneity within asset classes when mapping income flows to wealth. Our approach reduces bias in wealth estimates because wealth and rates of return are correlated. Overall, wealth is very concentrated: the top 1% holds as much wealth as the bottom 90%. However, the "P90-99" class holds more wealth than either group after accounting for heterogeneity. Relative to a top 0.1% wealth share of more than 20% under equal returns, we estimate a top 0.1% wealth share of [15%] and find that the rise since 1980 in top wealth shares falls by [half]. Top portfolios depend less on fixed income and public equity, depend more on private equity and housing, and more closely match the composition reported in the SCF and estate tax returns. Our adjustments reduce mechanical revenue estimates from a wealth tax and top capital income shares in distributional national accounts, which depend on well-measured estimates of top wealth. Though the capitalization approach has advantages over other methods of estimating top wealth, we emphasize that considerable uncertainty remains inherent to the approach by showing the sensitivity of estimates to different assumptions.</p><p>Written with Matthew Smith and Owen Zidar .</p><p><i>Finance Seminars at Caltech are funded through the generous support of The Ronald and Maxine Linde Institute of Economic and Management Sciences (lindeinstitute.caltech.edu).</i></p>Thu, 05 Mar 2020 16:00:00 -0800http://www.cms.caltech.edu/events/86147CMS-EE Partners Tech Talk: TBDclairer@caltech.edu (Claire Ralph)CMS-EE Partners Tech Talk<strong>Speaker(s):</strong> <br><strong>Location:</strong> Annenberg 105<br><p></p>Mon, 09 Mar 2020 12:00:00 -0700http://www.cms.caltech.edu/events/87006KNI Distinguished Seminar: 3D Laser Nanoprintingtkimoto@caltech.edu (Tiffany Kimoto)KNI Distinguished Seminar<strong>Speaker(s):</strong> Martin Wegener (Karlsruhe Institute of Technology (KIT))<br><strong>Location:</strong> Broad Center Rock Seminar Room<br><h4>Speaker: Martin Wegener, Karlsruhe Institute of Technology (KIT)</h4><h4>Talk Title: "3D Laser Nanoprinting"</h4><p></p><p><b>Abstract</b></p><p></p><p>In this talk, Dr. Wegener will give an introduction into laser based 3D printing on the micro- and nanoscale, describe the state-of-the-art, and compare it to other 3D additive manufacturing approaches. He will emphasize two future challenges: (i) scalable and faster 3D printing and (ii) multi-material 3D printing, including "4D" architectures. Application examples include micro-optical components, 3D metamaterials as meta-inks, scaffolds for biological cell culture, and 3D fluorescent security features.</p><p></p><p><b>Biography</b></p><p>After completing his Diplom and PhD in physics at Johann Wolfgang Goethe-Universität Frankfurt (Germany) in 1986 and 1987, respectively, he spent two years as a postdoc at AT&T Bell Laboratories in Holmdel (U.S.A.). From 1990-1995 he was professor (C3) at Universität Dortmund (Germany), since 1995 he is professor (C4, later W3) at Institute of Applied Physics of Karlsruhe Institute of Technology (KIT). Since 2001 he has a joint appointment as department head at Institute of Nanotechnology (INT) of KIT, since 2016 he is director at INT. From 2001-2014 he was the coordinator of the DFG-Center for Functional Nanostructures (CFN) at KIT. Since 2018 he is spokesperson of the Cluster of Excellence 3D Matter Made to Order. </p><p>His research interests comprise ultrafast optics, (extreme) nonlinear optics, optical laser lithography, photonic crystals, optical, mechanical, electronic, and thermodynamic metamaterials, as well as transformation physics. </p><p>This research has led to various awards and honors, among which are the Alfried Krupp von Bohlen und Halbach Research Award 1993, the Baden-Württemberg Teaching Award 1998, the DFG Gottfried Wilhelm Leibniz Award 2000, the European Union René Descartes Prize 2005, the Baden-Württemberg Research Award 2005, the Carl Zeiss Research Award 2006, the Hector Research Award 2008, the SPIE Prism Award 2014 for the start-up company Nanoscribe GmbH, the Stifterverband Science Award – Erwin-Schrödinger Prize 2016, and the Technology Transfer Prize of the German Physical Society (DPG) 2018. In 2014, 2015, 2016, 2017, and 2018, Clarivate Analytics listed him as "Highly Cited Researcher" (top 1%). He is Member of Leopoldina, the German Academy of Sciences (since 2006), Member of acatech, the National Academy of Science and Engineering (since 2019), Member of the Hector Fellow Academy (since 2013, presently also President), Fellow of the Max Plack School of Photonics (since 2019), Fellow of the Optical Society of America (since 2008), and Honorary Professor at Huazhong University of Science & Technology, Wuhan, China (since 2014).</p><p></p><p><a href="http://kni.caltech.edu/programs/kni-distinguished-seminar"><i>The KNI Distinguished Seminar Series</i></a> <i>is a new monthly series hosted by The Kavli Nanoscience Institute where eminent scientists and thinkers with strong yet varied backgrounds in nanoscience and nanotechnology share their expertise with the Caltech community. Seminars consist of a one-hour presentation, followed by a Q&A and light reception. The scopes of presentations may range from: recent outstanding scientific highlights/technological advancements, to innovative early-stage research developments, to broader cross-disciplinary topics that are relevant to nanoscience. Each seminar will be recorded and made available to the public via the</i> <a href="https://www.youtube.com/channel/UCWbGhTKjT6wv7T4sZ23E7WQ"><i>KNI's YouTube channel</i></a><i>.</i></p><p></p>Tue, 10 Mar 2020 16:00:00 -0700http://www.cms.caltech.edu/events/88045TCS+ Talk: TBAbjleung@caltech.edu (Bonnie Leung)TCS+ Talk<strong>Speaker(s):</strong> Thomas Steinke (IBM Almaden)<br><strong>Location:</strong> Annenberg 322<br><p><b>Abstract:</b> TBA</p>Wed, 11 Mar 2020 10:00:00 -0700http://www.cms.caltech.edu/events/88070Finance Seminar: Topic to be announcedsabrina@hss.caltech.edu (Sabrina Hameister)Finance Seminar<strong>Speaker(s):</strong> Eric D. Hilt (Wellesley College)<br><strong>Location:</strong> Baxter B125<br><p>Please check later for additional details</p><p><i>Finance Seminars at Caltech are funded through the generous support of The Ronald and Maxine Linde Institute of Economic and Management Sciences (lindeinstitute.caltech.edu).</i></p>Thu, 02 Apr 2020 16:00:00 -0700http://www.cms.caltech.edu/events/86252CMX Lunch Seminar: TBAjbrink@caltech.edu (Jolene Brink)CMX Lunch Seminar<strong>Speaker(s):</strong> <br><strong>Location:</strong> Annenberg 213<br><p></p>Wed, 15 Apr 2020 12:00:00 -0700http://www.cms.caltech.edu/events/86833CMX Lunch Seminar: TBAjbrink@caltech.edu (Jolene Brink)CMX Lunch Seminar<strong>Speaker(s):</strong> <br><strong>Location:</strong> Annenberg 213<br><p></p>Wed, 22 Apr 2020 12:00:00 -0700http://www.cms.caltech.edu/events/86834CMX Lunch Seminar: TBAjbrink@caltech.edu (Jolene Brink)CMX Lunch Seminar<strong>Speaker(s):</strong> Niles Pierce (Caltech)<br><strong>Location:</strong> Annenberg 213<br><p></p>Wed, 29 Apr 2020 12:00:00 -0700http://www.cms.caltech.edu/events/86835Innovation by Evolution: Bringing New Chemistry to Life: TBDmaria.cervantes@caltech.edu (Maria Cervantes)Innovation by Evolution: Bringing New Chemistry to Life<strong>Speaker(s):</strong> <br><strong>Location:</strong> Baxter Lecture Hall<br><p>Not satisfied with nature's vast catalyst repertoire, we want to create new protein catalysts and expand the space of genetically encoded enzyme functions. I will describe how we can use the most powerful biological design process, evolution, to optimize existing enzymes and invent new ones, thereby circumventing our profound ignorance of how sequence encodes function. Extending the capabilities and uncovering the mechanisms of these newly-evolved enzymes provide a basis for discovering biocatalysts for increasingly challenging reactions. These capabilities increase the scope of molecules and materials we can build using synthetic biology and move us closer to a sustainable world where chemical synthesis can be fully programmed in DNA.</p>Thu, 30 Apr 2020 16:00:00 -0700http://www.cms.caltech.edu/events/88027