| |||||||||
| Upcoming | |||||||||
 | |||||||||
 |
Wednesday, 05.22.13, 10 AM - 11 AM, Jacobs Hall, Room 2512 (Booker Conference Room) Topological Interference Management Syed Ali Jafar , UC Irvine W..more > | ||||||||
| Gone | |||||||||
| |||||||||
 |
Monday, 05.13.13, 2:00pm - 3:00pm, Jacobs Hall, Booker Room 2512 Networked Information Processing: New Compression, Processing and Control Paradigms for Networks Emrah Akyol, UC Santa Barbara Resource (energy, bandwidth, delay, storage) constraints are the key factors in designing the current and emerging communications, networking and signal processing applications. Such applications strongly motivate foregoing the convenience of digital communication and revisiting the potential benefits, and even necessity of analog communication, albeit in the modern context of source-channel coding. In the first part of the talk, I will present the major challenges and my research efforts in analog networking, ... | ||||||||
 |
Tuesday, 04.09.13, 11:00am - 12:00pm, Calit2, Atkinson Hall, Room 4004 Variable-Length Coding with Feedback: Finite-Length Analysis and Optimization Tsung-Yi Chen, UCLA This talk will show that the performance gap between these methods and the optimized variable-length coding scheme can be significant without a careful treatment in system design. | ||||||||
 |
Tuesday, 03.12.13, 11:00am - 12:00pm, Jacobs Hall, Room 2512, Henry G. Booker Conference Suite From Likelihood Filtering to Quantum Probabilities Hans-Andrea Loeliger, ETH Zurich The talk is about some recent ideas in statistical modeling and signal processing using state space models. | ||||||||
 |
Friday, 02.01.13, 3:30pm - 4:30pm, Jacobs Hall, Room 2512, Henry G. Booker Conference Suite Integration of Sensing, Communication, and Navigation in Mobile Networks Yasamin Mostofi, University of California, Santa Barbara In this talk, I develop a foundation for the integration of sensing, communication and navigation in mobile networks. | ||||||||
 |
Thursday, 01.10.13, 4:00pm - 5:00pm, AP&M Building, Room 6402 Near-Optimal Quantization and Encoding for Oversampled Signals Rayan Saab, Duke University A good A/D scheme allows for accurate reconstruction of the original object from its quantized samples. In this talk we investigate the reconstruction error as a function of the bit-rate, of Sigma-Delta quantization, a class of quantization algorithms used in the oversampled regime. | ||||||||
 |
Wednesday, 11.28.12, 2:00pm - 3:00pm, Jacobs Hall, Room 4309 Two Network Problems and Their Application to Streaming Codes Tracey Ho, Caltech We consider two established network-related problems, whose complete solutions are open. The first is a resource allocation problem that can be posed in a distributed storage context as follows. The problem is to store a unit size data object on a set of storage nodes such that the total amount of storage used does not exceed a given budget, and the probability of successful recovery is maximized under a given probabilistic model for node failure/accessibility. The second is network error correction coding for reliable communication over networks where arbitrary errors can occur on an unknown subset of links. We describe how some of our recent results on these two problems can be combined to design and analyze erasure correction coding for online streaming under probabilistic packet erasures. | ||||||||
 |
Monday, 11.19.12, 11 AM - 12 PM, Atkinson Hall, Room 4004 The Art of Gambling in a Team: Multi-player multi-armed bandits Rahul Jain, University of Southern California Multi-Armed bandits are an elegant model of learning in an unknown and uncertain environment. Such models are relevant in many scenarios, and of late have received increased attention recently due to various problems of distributed control that have arisen in wireless networks, pricing models on the internet, etc. We consider a non-Bayesian multi-armed bandit setting proposed by Lai & Robbins in mid-80s...more > | ||||||||
 |
Thursday, 11.15.12, 10 a.m - 11 a.m, Applied Physics & Mathematics Building (AP&M), Room 6402 A toy limit order book Elena Yudovina, University of Michigan I consider a Markov process inspired by a toy model of a limit order book. "Bid" and "ask" orders arrive in time; the prices are iid uniform on [0,1]. (I'll discuss some extensions.) When a match is possible (bid > ask), the highest bid and lowest ask leave the system. This process turns out to have surprising dynamics, with three limiting behaviours occurring with probability one...more > | ||||||||
 |
Thursday, 11.08.12, 2:00 PM - 3:00 PM, Jacobs Hall, Room 2512 (Booker Conference Room) Modeling Random Access in Underwater Acoustic Networks Paolo Casari, University of Padova, Italy T..more > | ||||||||
 |
Friday, 11.02.12, Atkinson Hall CWC Seminar If you would like to attend, please register by Wednesday 10/31 Morning presentations 8:30 Continental breakfast 9:00 Welcome 9:05 Content-centric message forwarding in ad-hoc networks, Rene Cruz, UCSD 9:33 Ultra-low-power radios for miniaturized wireless systems, Patrick Mercier, UCSD 10:01 Cloud mobile media: opportunities and challenges, Sujit Dey, UCSD 10:29 Break 10:44 Qualcomm Research: Radios and a whole lot more, Charles Bergan, Qualcomm 11:12 Real-time communication over unreliable wireless channels, P...more > | ||||||||
 |
Thursday, 11.01.12, 2pm - 3pm, Calit2 Atkinson Hall 4004 Lossy Compression for BigData: First Steps Dr. Thomas Courtade, Stanford University Two key challenges in fitting BigData problems into a lossy compression framework are (i) the selection of an appropriate distortion measure, and (ii) characterizing the performance of distributed systems. Inspired by real systems, like web search, which return a list of likely data entries indexed by likelihood, we study the "logarithmic loss" distortion function in a multiterminal setting, thus addressing both challenges. In particular, we characterize the rate-distortion region for two (generally open) multiterminal source coding problems when distortion is measured under logarithmic loss. In addition to the main results, I'll discuss applications to machine learning, estimation, and combinatorics. | ||||||||
 |
Friday, 10.26.12, 2:00pm, Jacobs Hall, Room 4309 Towards Computational Sensing through large number of Networked Sensors Prof. Lin Zhang, Dept. of Electronic Engineering, Tsinghua University Today, Internet users and networked sensors generate hundreds of gigabytes of data every minute, yet people are still feeling lost in an ocean of data, sometimes being starved of knowledge. In this talk, I will try to convince the audience, by sharing some experimental results in the systems the we developed and deployed in the past few years, that a surprisingly better understanding of the reality could be attained by performing state-of-art algorithms over large volume of data collected through networked sensors...more > | ||||||||
 |
Thursday, 10.25.12, 3:00 PM - 4:00 PM, Jacobs Hall, Room 2512 (Booker Conference Room) Three Open Problems in Network Communication Michael Langberg, The Open University of Israel In this talk I will discuss three natural open questions in the context of multi-source/ multi-terminal network communication via network coding. (a) What is the maximum loss in communication rate experienced from removing a single unit capacity edge from a given network? (b) What is the maximum loss in rate when insisting on zero error communication as opposed to vanishing decoding error? (c) What is the maximum loss in rate when comparing the communication of source information that is ``almost' independent to that of independent source information? Recent results including intriguing connections between the three questions will be presented...more > | ||||||||
 |
Thursday, 10.04.12, 10 a.m - 11 a.m, Applied Physics & Mathematics Building (AP&M), Room 6402 Queue-Size Scaling in Switched Networks Devavrat Shah , MIT, visiting Stanford University We consider a switched (queueing) network in which there are constraints on which queues may be served simultaneously; such networks have been used to effectively model input-queued switches, wireless networks and more recently data-centers. The scheduling policy for such a network specifies which queues to serve at any point in time, based on the current state or past history of the system...more > | ||||||||
 |
Tuesday, 08.14.12, 11:00 am - 12:00 pm, Jacobs Hall, Room 4309 Long Range Dependent Markov Models Barlas Oguz, Ph.D., University of California, Berkeley We discuss countable state Markov chains as a flexible class of models for long range dependent sources. We state sufficient conditions under which an instantaneous function of a long range dependent Markov chain has the same Hurst index as the underlying chain. We discuss several applications of the theorem in the fields of information theory, queuing networks, and finance. | ||||||||
 |
Friday, 06.15.12, 11:00am - 12:00pm, Atkinson Hall (Calit2), Room 4004 ITA Seminar: On q-ary Antipodal Matchings and Applications Dr. Gadiel Seroussi, HP Laboratories We define a q-ary antipodal matching to be a perfect matching in the bipartite graph with vertices corresponding to words of length m over the integer alphabet Q={0,1,...,q-1} wherein the left and right vertices are those with respective component sums greater and smaller than m(q-1)/2, and wherein two vertices are connected by an edge if one of the corresponding words dominates the other. We present two different constructions of efficiently computable q-ary antipodal matchings. We then show how such matchings can be used for encoding arbitrary data into n x n arrays over the alphabet Q all of whose row and column sums are at most n(q-1)/2. Such encoders might be useful for mitigating parasitic currents in a next generation memory technology based on crossbar arrays of resistive devices. (Joint work with Erik Ordentlich and Ronny Roth.) | ||||||||
 |
Wednesday, 05.23.12, 11 AM - 12 PM, Atkinson Hall, Room 4004 Error Correcting Codes for Distributed Control Ravi Teja Sukhavasi , Caltech Coding and information theory, the joint pillars upon which the telecommunications revolution has thrived, deal with how to reliably transmit information over unreliable channels. This is done by gathering large blocks of data and encoding them into larger blocks of data before transmitting, thereby achieving reliability at the expense of encoding/decoding delay...more > | ||||||||
 |
Monday, 04.30.12, 3:00 pm - 4:00 pm, Jacobs Hall, Room 2512, Henry G. Booker Conference Suite On Source-Channel Communication in Networks Jun Chen, McMaster University This talk is divided into two parts. In the first part of this talk, I will present several results on the optimality and approximate optimality of the source-channel separation architecture for lossy source coding in general networks. These results are shown without explicitly characterizing the achievable joint source-channel coding distortion region or the achievable separation-based coding distortion region. The second part of this talk is devoted to the problem of sending two correlated vector Gaussian sources over a bandwidth-matched two-user scalar Gaussian broadcast channel, where each receiver wishes to reconstruct its target source under a covariance distortion constraint. I will present a lower bound on the optimal tradeoff between the transmit power and the achievable reconstruction distortion pair. The derivation of this lower bound is based on a new bounding technique which involves the introduction of appropriate remote sources. Furthermore, it is shown that this lower bound is achievable by a class of hybrid schemes for the special case where the weak receiver wishes to reconstruct a scalar source under the mean squared error distortion constraint. This talk is based on joint work with Lin Song, Chao Tian, Suhas Diggavi, and Shlomo Shamai. | ||||||||
 |
Tuesday, 02.14.12, 10:30 a.m - 11:30 a.m, Jacobs Hall, Room 2512 (Booker Conference Room) Compressive Depth Acquisition Cameras: Principles and Demonstrations Vivek Goyal, MIT LIDAR systems and time-of-flight cameras use time elapsed from transmitting a pulse and receiving a reflected response, along with scanning by the illumination source or a 2D sensor array, to acquire depth maps. We introduce depth map acquisition with high spatial and range resolution using a single, omnidirectional, time-resolved photodetector and no scanning components. This opens up possibilities for 3D sensing in compact and mobile devices. Spatial resolution in our framework is rooted in patterned illumination or patterned reception. In contrast to compressive photography, the information of interest -- scene depths -- is nonlinearly mixed in the measured data. The depth map construction uses parametric signal modeling to achieve fine depth resolution and to essentially linearize the inverse problem from which spatial resolution is recovered. We have demonstrated depth map reconstruction for both near and medium-range scenes, with and without the presence of a partially-transmissive occluder. Our compressive depth acquisition camera (CoDAC) framework is an example of broader research themes of exploiting time resolution in optical imaging and identifying and exploiting structure in inverse problems. | ||||||||
 |
Monday, 01.30.12, 3:00 PM - 4:00 PM, Jacobs Hall, Room 2512 (Booker Conference Room) Algorithmic Phase Transitions Dimitris Achlioptas, UC Santa Cruz Constraint Satisfaction Problems are the common abstraction of numerous real-life settings, ranging from air-traffic design to protein folding. The ubiquity of CSPs presses forward a fundamental question: why is finding solutions to certain CSP instances exceptionally hard while often for, seemingly similar, instances it is quite easy? To study this phenomenon in a principled way we consider probability distributions over CSP instances formed by adding constraints one-by-one, uniformly and independently...more > | ||||||||
 |
Friday, 01.27.12, 11 AM - 12 PM, Atkinson Hall, room 4004 Load balancing for dynamically scalable web services Yi Lu, UIUC Web services have traditionally benefited from efficient load balancing. With multi-tenancy and the need for dynamic scalability in the cloud environment, existing hardware load balancers are expensive, difficult to scale and difficult to divide among users. Software load balancers are cost-efficient, scalable and divisible, but lack efficient load balancing algorithms...more > | ||||||||
 |
Wednesday, 01.25.12, 11 AM - 12 PM, Room 2512, Jacobs Hall (Booker Conference Room) An Exploration of Sparse Projections: Networks, Signals and Genetics Soheil Feizi, Graduate Student, MIT Using sparse projections can provide benefits in signal representation, reconstruction, and computation in different areas. For applications where signal sources are distributed across a noisy network, managing sparsity, whether naturally occurring or introduced, can be used to reconstruct/compute source signals in a computationally effective way. We introduce a joint source-channel-network coding scheme, which uses compressive sensing principles. Since there is no separation of channel, source and network coding, joint techniques are natural and this one allows for an integrated way to deal with these different aspects together. For applications such as cloud computing and functional compression, where computation, rather than reconstruction only, over a network is important, we introduce a sparse network design technique based on functional information flows. We conclude by presenting some applications of sparse representation in biological systems. In monitoring, our new techniques can considerably lower energy use in body area networks, even when compared to traditional compressive sensing. In motif discovery, we illustrate the effectiveness of our information-theoretically inspired techniques. | ||||||||
 |
Wednesday, 11.02.11, 11 AM - 12 PM, Jacobs Hall, Room 2512 (Booker Conference Room) New Channels for Underwater Acoustic Communication: The Particle Velocity Channels Ali Abdi, PhD, Associate Professor, NJIT, Newark The scalar component of the acoustic field, i.e., the pressure channel, has been extensively used for underwater acoustic communication. In recent years, vector components of the acoustic field, such as the three components of acoustic particle velocity, are suggested in our group for underwater communication. Consequently, one can use vector sensors for underwater communication. The small size of vector sensor arrays is an advantage, compared to pressure sensor arrays commonly used in underwater acoustic communication. This is because velocity channels can be measured at a single point in space. So, each vector sensor serves as a multichannel device. This is particularly useful for compact underwater platforms, such as autonomous underwater vehicles (AUVs). Our research efforts have focused on two closely-related problems: channel modeling and transceiver design. Channel modeling research aims at characterization of those aspects of acoustic particle velocity channels such as channel correlations, delay and Doppler spreads, etc., which determine the communication system performance. Transceiver design addresses optimal use of vector sensors and particle velocity for data modulation and demodulation, equalization, etc. (work supported by NSF). | ||||||||
 |
Friday, 10.14.11, 11 AM - 12 PM, Jacobs Hall, room 4309 Sampling Online Social Networks Athina Markopoulou, UC Irvine Online Social Networks (OSNs) have recently emerged as a new Internet killer-application and are of interest to a range of communities, ranging from computer science and engineering to social sciences. OSNs are widely studied today based on samples collected through measurement of publicly available information. In this talk, I will give an overview of our recent work on sampling online social networks...more > | ||||||||
 |
Friday, 07.22.11, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Room 6504 Wireless Channel Uncertainty in Relay-Assisted Communication and Distributed Detection Systems Azadeh Vosoughi , University of Rochester One of the main challenges in wireless communications is coping with channel uncertainty. Dealing with this uncertainty, and the limitations it imposes, is tightly related to the specific system and its application. In this talk, we consider two systems, namely a wireless bi-directional relay-assisted communication system and a wireless distributed detection system...more > | ||||||||
 |
Tuesday, 05.24.11, 1:30 - 2:30 pm, CMRR Auditorium Information-theoretic and algorithmic applications of quantum conditional mutual information Jon Yard, Los Alamos National Laboratory Conditional mutual information I(A;B|C) plays an important role throughout information theory. It is at least as important in quantum information, where the richer nature of quantum states presents new challenges and at the same time, offers the possibility for wider applicability in physics and the theory of computation. In this talk, which assumes no prior knowledge of quantum mechanics, I will present joint work (arXiv:1010...more > | ||||||||
 |
Monday, 05.09.11, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Booker Conference Room, 2512 Modeling and Characterization of Large-Scale Wi-Fi Traffic in Public Hot-Spots Dr. Amitabha Ghosh, Princeton University Server side measurements from several Wi-Fi hotspots deployed in nationwide network over different types of venues from small coffee shops to large enterprises are used to highlight differences in traffic volumes and patterns. We develop a common modeling framework for the number of simultaneously present customers. Our approach has many novel elements: (a) We combine statistical clustering with Poisson regression from Generalized Linear Models to fit a non-stationary Poisson process to the arrival counts and demonstrate its remarkable accuracy; (b) We model the heavy tailed distribution of connection durations through fitting a Phase Type distribution to its logarithm so that not only the tail but also the overall distribution is well matched; (c) We obtain the distribution of the number of simultaneously present customers from an M_t/G/1 queuing model using a novel regenerative argument that is transparent and avoids the customarily made assumption of the queue starting empty at an infinite past; (d) Most importantly, we validate our models by comparison of their predictions and confidence intervals against test data that is not used in fitting the models...more > | ||||||||
 |
Wednesday, 04.13.11, 11:00 a.m. - 12:00 p.m. , Jacobs Hall, Booker Conference Room, 2512 On channel polarization and polar codes Eren Sasoglu, Ecole Polytechnique Fédérale de Lausanne, Switzerland Arikan's polar codes achieve the symmetric capacity of binary-input memoryless channels at low encoding and decoding complexity. Early work on polar codes has shown that their underlying principle, i.e., polarization, is not restricted to binary-input channels. We discuss polarization for non-binary channels and sources, and for multiple-access channels...more > | ||||||||
 |
Wednesday, 04.06.11, 11:00 a.m. - 12:00 p.m. , Jacobs Hall, Booker Conference Room, 2512 Finite-Constellation Capacity of the Gaussian Channel Yihong Wu, Princeton University Denote by C_m(snr) the Gaussian channel capacity with signal-to-noise ratio snr and input cardinality m. We show that as m grows, C_m(snr) approaches C(snr) = 1/2 log(1 + snr) exponentially fast. Lower and upper bounds on the exponent are given as functions of snr. We propose a family of input constellations based on the roots of the Hermite polynomials that achieves exponential convergence, while quantization-based schemes are shown to achieve only polynomial convergence...more > | ||||||||
 |
Friday, 04.01.11, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Booker Conference Room, 2512 Data transmission: non-asymptotic fundamental limits Yury Polyanskiy, Princeton University Noise is an inalienable property of all communication systems appearing in nature. Such noise acts against the very purpose of communication, namely the delivery of data to its destination with minimal possible distortion. This creates a problem that has been addressed by various disciplines over the past century. In particular, information theory studies the question of the maximum possible rate achievable by an ideal system under certain assumptions regarding the noise generation and structural design constraints...more > | ||||||||
 |
Wednesday, 03.30.11, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Booker Conference Room, 2512 Network Interference Management via Interference Alignment Viveck Cadambe, UC Irvine Interference alignment has recently emerged as an effective technique to manage interference - the principal bottleneck of rates in wireless communication systems. In this talk, I will explore the principles and applications of the technique of interference alignment. The canonical model to study interference is the wireless interference network which has K mutually interfering users sharing for the same spectrum...more > | ||||||||
 |
Friday, 01.07.11, 12:00PM - 1:00PM, Atkinson Hall 3004 Utility Optimal Scheduling in Networks: Small Delay and No Underflow Longbo Huang, USC The recently developed Lyapunov optimization technique (commonly known as Backpressure / Max-Weight) is a powerful tool for solving a large class of stochastic network optimization problems. In this talk, we extend the theory in two directions: (i) We prove that dramatically improved delay is achievable with a simple Last-In-First Out (LIFO)-Backpressure rule, (ii) We generalize to "processing networks" where processing actions combine commodities of different queues to produce outputs, which involves a challenging "no underflow" constraint. In the first part of the talk, we show that the LIFO-Backpressure algorithm can achieve utility within epsilon of optimality (for any epsilon>0), with O([log(1/epsilon)]^2) average delay. This dramatically improves upon the previous O(1/epsilon) delay bounds, and results in 95-98% delay reduction in practical implementations. Remarkably, LIFO-Backpressure achieves... | ||||||||
 |
Wednesday, 11.03.10, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Booker Conference Room, 2512 Network Science for Wireless Communication: Information Dissemination, Mobility, and Resilience Edmund Yeh, Yale University Over the past decade, there has been a concerted effort to develop a network science for studying physical, biological, social, and information networks within a common framework. Of particular interest is the understanding of connectivity, information dynamics, and robustness in large-scale networks with spatial location and mobility. In this talk, we discuss a number of recent results from the application of network science ideas to mobile wireless communication...more > | ||||||||
 |
Wednesday, 08.11.10, 2:00 PM - 3:00 PM, Jacobs Hall, Booker Conference Room, 2512 Variable-Length Markov Modeling of Training Data Yields Optimal Universal Classification of Finite-Length Individual Test Sequences Jacob Ziv, Technion - Israel Institute of Technology Consider a classifier that observes an individual training sequence X and a test sequence Y of length N, over a finite alphabet. The classifier’s task is to decide whether the test sequence has the same features as those captured by the training sequence (which may be longer than N). It is assumed that the degree of similarity is measured in terms of the empirical average length of typical substrings (“patches”) that identically (or almost identically) appear in both sequences...more > | ||||||||
 |
Friday, 07.30.10, 11:00 a.m. - 12:00 p.m., Atkinson Hall (CALIT2), room 4004 Decomposing permutations by cost-constrained transpositions Olgica Milenkovic, University of Illinois at Urbana-Champaign We address the problem of finding the minimum decomposition of a permutation in terms of transpositions with non-uniform cost. Our investigation is motivated by three different applications. The first application pertains to sorting of genomic sequences, while the second application is related to a generalization of the notion of a chemical channel (also known as trapdoor channel)...more > | ||||||||
 |
Wednesday, 06.09.10, 11:00 a.m. - 12:00 p.m. , Jacobs Hall, Booker Conference Room, 2512 Interactive distributed source coding for network function computation Nan Ma, Boston University Today, blocklength, rate, SNR, frequency, quantizer resolution, and network size are well-studied and recognized resources for communication in information theory. A relatively less well recognized and less understood resource for communication and computation is interaction. Interaction as a resource becomes particularly valuable in the context of distributed function computation where it may be necessary for nodes to exchange information bidirectionally, perform computations, and harness the structure of the functions, statistical-dependencies, and the network topology to maximize the overall computation-efficiency as opposed to only generating, receiving, and forwarding data...more > | ||||||||
 |
Wednesday, 05.05.10, 11:00 a.m. - 12:00 p.m., Jacobs Hall, Booker Conference Room, 2512 The Value of Volatile Resources in Electricity Markets Sean P. Meyn, University of Illinois at Urbana-Champaign While renewable resources most certainly provide environmental benefits, and also help to meet aggressive renewable energy targets, their deployment has pronounced impacts on system operations. There is an acute need to understand these impacts in order to fully harness the benefits of renewable resource integration. In this paper we focus on the integration of wind energy resources in a multi-settlement electricity market structure. We study the dynamic competitive equilibrium for a stochastic market model and obtain closed form expressions for the supplier and consumer surpluses. Numerical results based on these formulae show that the value of wind generation to consumers falls {dramatically} with volatility. In fact, we can establish thresholds for the coefficient of variation beyond which the value of wind is questionable. These findings can help guide the integration of renewables in future electricity markets. | ||||||||
 |
Thursday, 04.29.10, 2:00PM - 3:00PM, EBU-I, Room 2512 Henry G. Booker Conference Suite Mutual Information, Relative Entropy, and the Relationship Between Causal and Non-Causal Mismatched Estimation in AWGN Channels Tsachy Weissman, Stanford A continuous-time process with distribution P is observed through an Additive White Gaussian Noise channel, at a given signal-to-noise ratio (SNR), and is estimated by an estimator that would have minimized the mean-square error if the process had distribution Q. We show that the causal filtering mean-square error (MSE) achieved at SNR level snr is equal to the average value of the noncausal (smoothing) MSE achieved with a channel whose SNR is chosen uniformly distributed between 0 and snr. Emerging as the bridge for equating these quantities are mutual information and relative entropy. Our results build on and extend those of [Duncan 1970], [Guo, Shamai and Verdu 2005], and [Verdu 2009] in ways that will be explained. I will also present extensions that accommodate the presence of feedback and discuss implications on minimax estimation. | ||||||||
 |
Wednesday, 04.28.10, 1:30 p.m. - 2:30 p.m. , Atkinson Hall (CALIT2), room 6004 Matrix completion: fundamental limits and efficient algorithms Sewoong Oh , Stanford University A number of data sets are naturally described in matrix form. Examples range from micro-arrays to collaborative filtering data. In many of these examples, Singular Value Decomposition (SVD) is a powerful and widely-used technique to construct a low-rank approximation, thus achieving a great dimensionality reduction. However, when we have incomplete data, SVD is strictly sub-optimal...more > | ||||||||
 |
Friday, 04.23.10, 11:00AM - 12:00PM, 4004 Atkinson Hall Opportunistic Routing in Wireless Networks with Congestion Diversity Tara Javidi, UCSD Opportunistic routing for multi-hop wireless networks has seen recent research interest to overcome deficiencies of traditional routing. First, we, briefly, cast opportunistic routing as a Markov decision problem (MDP)and introduce a stochastic variant of distributed Bellman-Ford which provides a unifying framework for various versions of opportunistic routing such as SDF, GeRaF, and EXOR. In the second part of the talk, we touch upon the issue of congestion and throughput optimality by contrasting the opportunistic MDP-based schemes with back-pressure schemes. Inspired by the properties of backpressure algorithm, we propose a modification of the MDP framework to account for congestion and arrive at a throughput-optimal policy, aka ORCD, that exhibits significant delay improvements over the existing candidates in the literature. In the process of proving the throughput optimality of ORCD, we introduce a new Lyapunov function construction which characterizes a large class of throughput optimal policies. The proposed class includes backpressure and ORCD as simple special cases. | ||||||||
 |
Wednesday, 03.17.10, 11:00 AM - 12:00 PM, Atkinson Hall 4004 Large wireless networks: fundamental limits and design issues Paolo Minero, UCSD In this talk, we demonstrate how fundamental questions in large wireless networks can be addressed by applying methods from information theory, physics, networking and control. We focus on three examples of emerging systems architecture. First, we investigate the maximum achievable throughput in a wireless ad-hoc network. By combining Maxwell's physics of wave propagation and Shannon's theory of information, and departing from idealistic stochastic channel models for signal propagation, we derive an upper bound to the law that determines the scaling of throughput with the population size of the network, and conclude that the scaling achieved by multi-hop communication is optimal in any constant density wireless network...more > | ||||||||
 |
Wednesday, 02.17.10, 1:30 PM - 2:30 PM, 6004 Atkinson Hall On the Duality Between Slepian-Wolf Coding and Channel Coding Jun Chen, McMaster University Slepian-Wolf coding, also known as distributed lossless source coding, is of fundamental importance for many emerging applications. In this talk, we will discuss the intimate connections between Slepian-Wolf coding and channel coding. We show that there exist two different dualities between Slepian-Wolf coding and channel coding: type-level duality and linear codebook-level duality. These two dualities together provide a comprehensive picture of Slepian-Wolf coding and clarify many subtle differences between linear block codes, fixed-rate nonlinear codes, and variable-rate codes. The implication of this work on Slepian-Wolf code design will also be discussed. | ||||||||
 |
Friday, 12.11.09, 11:00 - 12:00, 4004 Atkinson Hall On information theoretic network security Tracey Ho, Caltech Information theoretic coding techniques can be used to provide security against data modification and eavesdropping in networks. Coding redundant information across multiple network links enables reliable communication over a network even when individual links cannot be made reliable, as in the case of adversarial data modification, for which it is not sufficient to do error correction on a link by link basis. Coding the source message together with random keys enables information theoretically secret communication over networks where a subset of links can be wiretapped. | ||||||||
 |
Friday, 11.13.09, 3:00 PM - 4:00 PM, EBU2, Room 479 Understanding Implicit Communication in Distributed Control Pulkit Grover, UC Berkeley In distributed systems, control actions often serve a dual purpose -- minimizing the immediate control cost, and communicating relevant information to help other controllers. Unfortunately, though this “implicit communication” is ubiquitous, it also seems to make such problems hard. General communication problems have been addressed quite successfully using information theory...more > | ||||||||
 |
Wednesday, 11.11.09, 2:00 PM - 3:00 PM, 4004 Atkinson Hall Coding for online adversaries Michael Langberg, The Open University of Israel In this talk we consider the communication of information in the presence of an "online" or "causal" adversarial jammer. In the setting under study, a sender wishes to communicate a message to a receiver by transmitting a codeword x=(x_1,...,x_n) symbol-by-symbol over a communication channel. The (malicious) adversarial jammer can view the transmitted symbols x_i one at a time, and can change up to a p-fraction of them. However, the decisions of the jammer must be made in an online (or causal) manner. Namely, for each symbol x_i the jammer's decision on whether to corrupt it or not (and on how to change it) must depend only on x_j for j <= i. This is in contrast to the ``classical' adversarial jammer which may base its decisions on its complete knowledge of the codeword x. | ||||||||
 |
Wednesday, 11.04.09, 1:00 PM - 2:00 PM, 4004 Atkinson Hall Network Compress-Forward Young-Han Kim, Electrical and Computer Engineering, UCSD In this talk, we extend the compress-forward coding scheme to a general network with multiple sources and multiple destinations, demonstrating that it provides a robust and scalable building block for relaying. In particular, we show that the network compress-forward coding scheme extends all previously known results on relay networks (except for decode-forward), including multicast network coding over noiseless networks by Ahlswede, Cai, Li, and Yeung, coding for wireless relay networks and deterministic networks by Avestimehr, Diggavi, and Tse, coding for wireless erasure networks by Dana, Gowaikar, Palanki, Hassibi, and Effros, and coding for linear erasure networks by Smith and Vishwanath. The key idea behind network compress-forward is message repetition coding in which the source transmits a single message over multiple transmission blocks. | ||||||||
 |
Wednesday, 10.28.09, 2:00 PM - 3:30 PM, 4004 Atkinson Hall Online Learning and Drifting games Yoav Freund, CSE Department, UCSD Online learning is an approach to statistical inference based on the idea of playing a repeated game. Most popular online learning strategies are based on exponential weights. In this talk I will present a different family of games, called drifting games, and show the relationship between them and online learning games. A new online learning algorithm is derived and analyzed using drifting games. This algorithm can also be analyzed in the context of continuous time, revealing a tight connection between these games and Brownian motion. | ||||||||
 |
Friday, 10.23.09, 11:00 - 12:00, 4004 Atkinson Hall Convergence Results for Ant Routing Algorithms via Stochastic Approximation Punyaslok Purkayastha, University of Maryland, College Park We consider a class of routing algorithms for communication networks called Ant-Based Routing Algorithms (ARA), that are inspired by experimental observations of ant colonies in nature. It was found that ant colonies are able to "discover" the shorter of two paths to a food source by laying and following "pheromone trails". This idea has been adapted for purposes of routing in communication networks, whereby probe packets (analogues of ants) are used to explore the network and collect measurements of path delays...more > | ||||||||
 |
Friday, 10.16.09, 11:00 a.m. - 12:00 p.m., Atkinson Hall (CALIT2), room 4004 Learning algorithms and Markov chain computations Vivek Borkar, School of Technology and Computer Science,Tata Institute of Fundamental Research This talk will describe potential applications of some reinforcement learning algorithms originally developed for policy evaluation for certain Markov decision problems, for some Markov chain computations. Notably, computation of stationary expectations and stationary distributions will be considered. Connections with Monte Carlo and numerical schemes will be highlighted and some possible acceleration schemes will also be discussed. | ||||||||
 |
Monday, 10.12.09, 4:00 p.m. - 5:00 p.m., EBU1, room 4309 Information Theory meets Game Theory on the Interference Channel Randall Berry , Northwestern University As wired and wire-line communication networks migrate to more open models (e.g.~open spectrum access), it is becoming increasingly important to understand the interaction of different users, who may not have an incentive to cooperate with each other. Such questions are naturally studied using game theory. Here, we consider a canonical example of such a problem, namely a game among users sharing a Gaussian interference channel. Previous work on such "interference games" places restrictions on the encoding and decoding strategies of the users and thus are not truly information theoretic in nature. In this talk we discuss a general formulation for interference games that allows users to employ any encoding and decoding strategy. We discuss the solution of these games for the two-user linear deterministic interference channel and then partially extend these results to Gaussian interference channels. This is joint work with David Tse at Berkeley. | ||||||||
 |
Wednesday, 08.26.09, 2:00 PM - 3:00 PM, Atkinson Hall, Room 4004 On the optimality of universal classifiers for finite-length individual test sequences Dr. Jacob Ziv, Technion-Israel Institute of Technology An empirical informational divergence (relative entropy) between two individual sequences has been introduced in [1]. It has been demonstrated that if the two sequences are independent realizations of two finite-order, finite alphabet, stationary Markov processes, the proposed empirical divergence measure (ZMM), converges to the relative entropy almost surely...more > | ||||||||
 |
Wednesday, 07.22.09, 11:00 - 12:00, 3004 Atkinson Hall Collaboration, Power and Stability Yoram Bachrach, Microsoft Research, Cambridge, UK I will consider models of cooperation in multi-agent settings, with or without strategic behaviour. Cooperation is easier in settings where the agents are not self interested. In collaborative filtering, agents attempt to recommend an information item, based on ratings of similar users. I will discuss a sketching method that allows storing much less information than the full ratings, and still achieve high quality recommendations. I will then consider several models of cooperation in multiagent systems from a computational game theoretic perspective. I will show how to compute power indices, which reflect how much "real power" an agent has in such environments. The common thread of both these techniques are randomized algorithms and a statistical analysis of the number of required samples. | ||||||||
 |
Friday, 06.19.09, 12:00 - 1:00, 4004 Atkinson Hall Reliable Relaying with Uncertain Knowledge Jiwoong Lee, University of California, Berkeley The motivation for this talkis to analyze the effect of information uncertainty on the design and performance of protocols. The talk considers two types of situations. The first is when different nodes in the network have bounded knowledge about what other nodes know. The second, called common knowledge about inconsistent beliefs, is when the information is inconsistent but everyone knows it. Situations of bounded or inconsistent information arise naturally in networks because the state of these systems changes and nodes take time to learn of those changes. | ||||||||
 |
Monday, 06.08.09, 12:00 - 1:00, 4004 Atkinson Hall Surprisal in real-time human language comprehension and production Roger Levy, Department of Linguistics, University of California, San Diego Probabilistic modeling has revolutionized computational linguistics in the last fifteen years. In this talk, I describe how a fundamental quantity in probability theory -- surprisal, or self-information -- is starting to illuminate our understanding of real-time human language processing as well. The talk covers two fundamental issues, one each in language comprehension and production: what determines the difficulty of comprehending a given word in a given sentence, and what factors influence the choice that a speaker makes when it is possible to express a meaning more than one way? The first half of the talk covers models and experimental results in language processing that show how probabilistic expectations can be a stronger determinant of comprehension difficulty than more traditional measures of difficulty based on memory load...more > | ||||||||
 |
Tuesday, 05.26.09, 12:00 - 1:00, 4004 Atkinson Hall Coding Techniques for Distributed Storage Distributed storage schemes for Data centers and peer-to-peer networks often use erasure coding to introduce redundancy for robustness. We introduce novel network coding techniques that can surprisingly reduce the communication required to maintain a storage system compared to standard Reed-Solomon codes used in current architectures. We present novel information theoretic performance bounds and explicit network codes that outperform known storage coding techniques. We show the connections of the storage repair problem to matroid theory and a computer search approach to finding optimal storage codes. | ||||||||
 |
Monday, 05.18.09, 4:00 - 5:00, Atkinson Hall Room 4004 A Stochastic Control Viewpoint on `Posterior Matching'-style Feedback Communication Schemes This paper re-visits Shayevitz & Feder's recent `Posterior Matching Scheme', a deterministic, recursive, capacity-achieving feedback encoding scheme for memoryless channels. We here consider the feedback encoder design problem from a stochastic control perspective. The state of the system is the posterior distribution of the message given current outputs of the channel. The per-trial reward is the average `reduction in distance' of the posterior to the target unit step function. We show that the converse to the channel coding theorem with feedback upper bounds the optimal reward, and that the posterior matching scheme is an optimal policy. We illustrate that this `reduction in distance' symbolism leads to the existence of a Lyapunov function on the Markov chain under this optimal policy, which leads to demonstration of achievability for all rates less than capacity. | ||||||||
 |
Monday, 05.11.09, 12:00 - 1:00, Atkinson Hall Room 4004 Information theoretic generation of secret keys / Evaluation of Marton’s Inner Bound for the General Broadcast Channel Amin Gohari, University of California, Berkeley The talk consists of two main parts. In the first part of the talk, I will present our new bounds for two important models defined in the context of information-theoretic security. In the second part of the talk, I will introduce a non-Caratheodory type tool for proving cardinality bounds, and then apply the tool to derive cardinality bounds on the auxiliaries of Marton’s inner bound for the general broadcast channel. | ||||||||
 |
Monday, 05.04.09, 11:45 - 12:45, Atkinson Hall Room 4004 Spectral Clustering of Surfaces Ery Arias-Castro, University of California, San Diego A typical generative model for surface clustering assumes that each cluster is the result of sampling a number of points in the neighborhood of a surface. In this setting, we provide theoretical guaranties for the pairwise spectral clustering technique of Ng, Jordan and Weiss (NIPS'01). We also introduce a new multi-way spectral clustering method based on local linear (or higher order) approximations, for which we provide theoretical guaranties as well...more > | ||||||||
 |
Monday, 04.20.09, 12:00 - 1:00, Atkinson Hall Room 4004 Subspace Pursuit for Compressive Sensing and Its Applications Wei Dai, University of Illinois at Urbana-Champaign Compressive sensing (CS) is a technique for acquiring sparse signals efficiently. It has recently received significant attention, due to its large potential for practical applications in signal processing, statistics, and wireless communications, and due to its provable reconstruction performance guarantee via polynomial complexity algorithms. In this talk, we will first give a brief tutorial on CS and then focus on a greedy algorithm for signal reconstruction in CS, termed the subspace pursuit (SP) algorithm. The algorithm has two important characteristics: reconstruction accuracy of the same order as that of the benchmark optimization methods, and low computational complexity especially when the unknown signal is sufficiently sparse. Because of the inherent simple structure, the SP algorithm can be extended to different scenarios. We will see how to modify or apply this algorithm to accommodating quantization effects, solving matrix completion problems, and improving the existing interference cancellation techniques in wireless communications. | ||||||||
 |
Monday, 04.13.09, 12:00 - 1:00, TBA Automatic Verification of Data-centric Business Processes Alin Deutsch, University of California, San Diego The talk presents a study of business process systems that are centered around "business artifacts", or simply "artifacts". This approach focuses on data records, known as artifacts, that correspond to key business-relevant objects, and that flow through a business process specified by a set of services. The artifact-centric approach has been introduced by IBM, and has enabled significant improvements to the operations of medium- and large-sized businesses. Artifacts carry attribute records and internal state relations, that services can consult and update. In addition, services can access an underlying database and can introduce new values from an infinite domain, thus modeling external inputs or partially specified processes described by pre-and-post conditions. | ||||||||
 |
Wednesday, 03.11.09, 11:30 - 12:30, Atkinson Hall, Room 6006 An information-theoretic view of early perception (or how nature discovered mutual information long before Shannon) Nuno Vasconcelos, Department of Electrical and Computer Engineering, UCSD Fifty years ago, Hubel and Wiesel revolutionized neuroscience through the measurement of neural responses in the primary visual cortex of the cat. They uncovered a computational architecture, based on simple and complex cells, which has been accepted as a good model for early perception over the decades. I will review some of our work on a new interpretation of these neural computations as the measurement of certain types of mutual information, for perceptual stimuli that follow statistical distributions commonly found in the natural world. This has a number of interesting consequences for both the understanding of biological perception and the development of new algorithms for computer vision. I will review some examples in the area of visual saliency, object recognition, and visual tracking. | ||||||||
 |
Monday, 03.02.09, 11:30 AM - 12:30 PM, Atkinson Hall, Room 4004 Haplotype assembly Vineet Bafna, University of California, San Diego The availability of high density SNP chips has empowered whole-genome association scans for many common diseases. However, current genotyping methods do not reveal haplotypes: the combination of alleles at neighboring SNPs on a single chromosome that tend to be inherited together. Knowledge of haplotypes is important for fine-scale mapping of disease-related variants, and understanding the role of different evolutionary mechanisms-meiotic recombination, positive selection-in shaping human genetic variation. Statistical methods, based on population genotypes analysis can be used, but are limited for long range haplotyping. Parallel advances in sequencing technologies now allow sequencing of individual genomes; they also enable haplotyping. In this talk, I will describe combinatorial and stochastic (Markov Chain Monte Carlo) algorithms for reconstructing long and accurate haplotypes from whole genome sequence data for an individual (J. Craig Venter). While the overall method is a heuristic one, relying on computing cuts in an associated graph, it is motivated by a theoretical analysis of the mixing time for representative markov chains, where we show that two similar markov chains have very different mixing properties. Experimental results on the Venter data, and simulations validate the power of our approach to haplotype assembly. | ||||||||
 |
Wednesday, 02.18.09, 11:30 - 12:30, Atkinson Hall, Room 6006 Graphical Models for Networks Sujay Sanghavi, Electrical and Computer Enigneering, Purdue University Graphical models combine probability theory and graph theory into a powerful formalism, facilitating the use of graph algorithms to simplify inferring about, sampling from, and learning of probability distributions. They have found applications in statistical physics, multivariate signal processing, machine learning, communications etc. In this talk we present a broad-based application of this formalism to networks. The talk will be made broadly accessible; in particular, it does not require pre-existing familiarity with graphical models. | ||||||||
 |
Wednesday, 01.28.09, 11:30 - 12:30, Atkinson Hall, Room 6006 Invertible Extractors and Wiretap Protocols Mahdi Cheraghchi, Laboratory of Algorithmic Mathematics, Swiss Federal Institute of Technology at Lausanne (EPFL) A wiretap protocol is a pair of randomized encoding and decoding functions such that knowledge of a bounded fraction of the encoding of a message reveals essentially no information about the message, while knowledge of the entire encoding reveals the message using the decoder. We study the notion of efficiently invertible extractors and show that a wiretap protocol can be constructed from such an extractor. We will then construct invertible extractors for symbol-fixing and affine sources and apply them to create wiretap protocols with asymptotically optimal trade-offs between their rate (ratio of the length of the message versus its encoding) and resilience (ratio of the observed positions of the encoding and the length of the encoding). We will then apply our results to create wiretap protocols for challenging communication problems, such as active intruders who change portions of the encoding and the wiretap problem in network coding. | ||||||||
 |
Wednesday, 01.21.09, 11:30 AM - 12:30 PM, Atkinson Hall, Room 6006 The Liar's Game with a List, and Error Correction with Feedback Ofer Shayevitz, UCSD Consider the following game played by Alice and Bob. Alice picks one of M objects, and Bob asks n yes/no questions trying to learn which one. Alice is allowed to lie at most t times, and Bob is declared the winner if he is able to provide a list of L objects containing the one selected by Alice. Does there exist a strategy which guarantees that Bob can always win? This question is equivalent to the problem of binary error correction with feedback under list-of-L decoding, and was extensively studied in the past for the unique decoding case of L=1. | ||||||||
 |
Wednesday, 01.14.09, 11:30 AM - 12:30 PM, Atkinson Hall, Room 4004 IEEE 802.11 is Good Enough to Build Wireless Multi-Hop Networks Apoorva Jindal, University of Southern California We formally establish that IEEE 802.11 yields exceptionally good performance in the context of wireless multi-hop networks. A common misconception is that existing acceptable CSMA-CA random access schemes like IEEE 802.11 yield unfair and inefficient rates in wireless multi-hop networks. This misconception is based on works which study IEEE 802.11-scheduled multi-hop networks with TCP or in saturation conditions both of which grossly underutilize the available capacity that IEEE 802.11 provides, or use topologies which cannot occur in practice due to physical layer limitations. | ||||||||
 |
Thursday, 08.28.08, 12:00 PM, CALIT2, Room 4004/06 On Universal Coding for Parallel Gaussian Channels Maryam Modir Shanechi , MIT Two classes of approximately universal codes are developed for parallel Gaussian channels whose state information is not available at the encoder. Both architectures convert the channel into a set of scalar additive white Gaussian noise (AWGN) channels to which good AWGN base codes can be applied, and both are layered schemes used in conjunction with successive interference cancellation to keep decoding complexity low...more > | ||||||||
 |
Tuesday, 06.03.08, 3:00 PM - 4:00 PM, CSE 4140 Average-case hardness for NP Andrej Bogdanov, Tsinghua University The gold standard of difficulty for a computational problem is NP hardness.However, NP hard instances of problems are quite atypical. In practice, and in many theoretical settings as well, one sometimes adopts the view that NP hard instances are so rare that they present no serious obstacle to the design of efficient algorithms. Yet experience tells us that some NP problems are intractable even on typical instances...more > | ||||||||
 |
Monday, 06.02.08, 12:00 PM, CALIT2, Room 3004 The Geometry of Ad Hoc Networks and its Impact on Performance Martin Haenggi , University of Notre Dame The node distribution in ad hoc and sensor networks is typically modeled as a stochastic point process. Due to its analytical tractability, the (homogeneous) Poisson point process (PPP) is widely popular. We give an overview of interference and outage results for the PPP, and we present an approach to extend these to more general point processes. Next we show how fading can be incorporated in the point process, which leads to a geometric interpretation of fading that permits a convenient characterization of single-hop connectivity and transport capacity...more > | ||||||||
 |
Monday, 04.21.08, 11:00 AM - 12:00 PM, EBU-1, 6504 Testing Low Degree Polynomials over Small Prime Fields Anindya Patthak , UC Riverside Polynomials are very important objects in mathematics and theoretical computer science. Every map from a finite field to finite field is a polynomial map. Low degree polynomials are more important for small complexity and for better error correction capability. In this talk, I will present how to test whether a given function is a low degree polynomial over prime fields by querying randomly at few points...more > | ||||||||
 |
Friday, 04.18.08, 11:00 AM - 12:00 PM, EBU 1, Room 6504 Efficient Algorithms for Active Learning Claire Monteleoni , UC San Diego The rapidly increasing abundance of data generated by internet transactions, satellite measurements, and environmental sensors, among other sources, creates new and urgent challenges for machine learning. My work on machine learning theory and algorithms is motivated by the problems posed by real-world data. This talk will focus on providing efficient algorithms for active learning, a rich model for learning when labels are missing...more > | ||||||||
 |
Tuesday, 04.15.08, 11 - 12, EBU1, Room 2512 Robust architectures for next generation communication systems Anand Sarwate, University of California, Berkeley Researchers have developed a good understanding of the communication systems of the past decades by using random noise models. This has shaped code design as well as overall system architectures. In this talk, I will argue that these models are not appropriate for the challenges of the future systems, such as cognitive radio, ad-hoc networks, and sensor networks...more > | ||||||||
 |
Monday, 04.07.08, 11:00 AM - 12:00 PM, EBU-I 6504 The posterior matching principle for optimal communication with feedback Ofer Shayevitz, Tel-Aviv University Feedback cannot increase the capacity of a memoryless channel. Nevertheless, feedback can sometimes significantly simplify the transmission scheme, so that capacity is achieved essentially without coding. This was demonstrated by the celebrated examples of Horstein (1963) for the binary symmetric channel (extendable to any DMC) and Schalkwijk-Kailath (1966) for the additive white Gaussian noise channel...more > | ||||||||
 |
Monday, 03.31.08, 11:00 AM - 12:00 PM, EBU1, Room 6504 Information flow over wireless networks: a deterministic approach Salman Avestimehr , UC Berkeley How does information flow over wireless networks? Answer to this basic question is one of the most challenging problems in the field of wireless network information theory. From a practical point of view, the answer to this question will have a significant impact on the architectural design of future wireless systems. So far, the majority of research done in this area has been based on the classical additive Gaussian noise model for wireless channels...more > | ||||||||
 |
Wednesday, 02.27.08, 10.30am - 11.30am, 4004/4006 (4th floor Atkinson) Computing on Streams: New Results and Directions Andrew McGregor, UCSD Over the last ten years, the data stream model has become an active topic of research. I think there are two main reasons for this. First, it is a model that is applicable to a range of practical problems such as monitoring network traffic, query-planning in databases, designing I/O efficient algorithms for massive data sets, and aggregation in sensor networks...more > | ||||||||
 |
Friday, 11.09.07, 1:30 PM - 2:30 PM, EBU 1, Room 4307 Finding low-rank matrices via convex optimization Maryam Fazel, Caltech In many engineering applications, notions such as order, complexity, or dimension of a model or design can be expressed as the rank of an appropriate matrix. If the set of feasible models or designs is convex, choosing the simplest model can be cast as a Rank Minimization Problem. For example, a low-rank matrix could correspond to a low-order controller for a plant, a low-degree statistical model for a random process, or an embedding in a low-dimensional space...more > | ||||||||
 |
Monday, 10.22.07, 11 AM - 12 PM, Calit2, Room 4004 Degrees of Freedom and O(1) Capacity of Wireless Interference Networks Syed Ali Jafar , UC Irvine The talk will present new insights into the capacity of fully connected wireless networks with finite number of nodes through capacity approximations that are accurate to within a bounded constant for all SNR. While the best known outerbound for the K user interference channel states that there cannot be more than K/2 degrees of freedom, it has been conjectured that in general the constant interference channel with any number of users has only one degree of freedom...more > | ||||||||
 |
Friday, 08.17.07, 2PM - 3PM, Calit2, Rm 4004 Capacity Approaching, Efficiently Decodable Lattice Codes Meir Feder, Tel Aviv University Lattice codes can achieve the capacity of the additive white Gaussian noise channel. However, the specifiic proposed lattice codes were either of finite block length, or were of random structure and require complex decoding to achieve good performance. On the other hand, in the recent years coding scheme that attain capacity and are decoded efficiently were proposed for finite alphabet channels, the most notable examples are Low Density Parity Check Codes (LDPC) and turbo codes...more > | ||||||||
 |
Thursday, 07.26.07, 10 - 11, Calit2 Rm. 4004 On universal compression and classification of individual two-dimensional arrays Jacob Ziv, The Technion, Israel A common approach to data compression of two-dimensional arrays is to first scan the array, thus mapping it into a one-dimensional sequence, and then apply a data-compression algorithm for sequences. It has been demonstrated that for individual arrays, a space-filling Peano-Hilbert scan that is followed by LZ data-compression algorithm for individual sequences is an asymptotically optimal two-stage scheme...more > | ||||||||
 |
Tuesday, 05.08.07, 1PM, TBD Clustering using Correlation and Independence Kamalika Chaudhuri, UC Berkeley Clustering, a method of finding structure in unlabelled data by grouping the data points into few groups based on a similarity measure, has many applications in AI, Physics and Biology. A simple theoretical model that captures clustering is the problem of learning mixtures of distributions. In this setting, one is given sample points generated from a mixture of T distributions of a certain type, and the goal is to recover these distributions and classify the points correctly...more > | ||||||||
 |
Monday, 05.07.07, 11:30, EBU-1 6504 Cooperative Diversity: Protocols and Analysis Kambiz Azarian Yazdi, University of Notre Dame In this talk, we overview several approaches to designing cooperative transmission schemes. In particular, we differentiate between the relay, broadcast (down-link), and multiple-access (up-link) scenarios and propose schemes that efficiently utilize the cooperative diversity available in the channel. We do so by re-examining the two basic cooperation techniques (i...more > | ||||||||
 |
Friday, 01.19.07, 2PM - 3PM, Calit 5004 Role of noisy feedback in communication Young-Han Kim, UCSD The role of perfect feedback in communication is relatively well understood. Feedback can dramatically reduce the probability of error, decrease communication delay, and simplify the system design. For example, the celebrated Schalkwijk-Kailath coding scheme achieves the capacity of additive white Gaussian noise channel with exponentially improved reliability...more > | ||||||||
 |
Tuesday, 12.05.06, 11 - 12, CSE 4217 Estimation of probability distributions with maximum entropy -- incorporating generalized regularizations and modeling species habitats Miroslav Dudik, Princeton University Maximum entropy (maxent) approach, equivalent to maximum likelihood, is a widely used method for estimating probability distributions. However, when trained on small datasets, maxent is likely to overfit. Therefore, many smoothing techniques were proposed to mitigate overfitting. In this talk, I will present a unified treatment for a large and general class of smoothing techniques including L1 and L2 regularization...more > | ||||||||
 |
Monday, 12.04.06, 11 - 12, CSE 4217 Convex Repeated Games and Fenchel Duality Shai Shalev-Shwartz, The Hebrew University Several problems arising in machine learning, such as online learning and boosting, can be modeled as a convex repeated game. A convex repeated game is a two players game which is performed in a sequence of consecutive trials. At each trial of the game, the first player predicts a vector from a predefined set and then the second player responds with a loss function over the set...more > | ||||||||
 |
Friday, 12.01.06, 11 AM - 12 PM, EBU-1 6504 Slepian-Wolf Coding over Broadcast Channels Ertem Tuncel, UC Riverside We discuss reliable transmission of a discrete memoryless source over a discrete memoryless broadcast channel, where each receiver has side information (of arbitrary quality) about the source unknown to the sender. When there are two receivers, the optimum coding strategy using separate and stand-alone source and channel codes is to build two independent binning structures and send bin indices using degraded message sets through the channel, yielding a full characterization of achievable rates...more > | ||||||||
 |
Monday, 10.30.06, 1:30 - 2:30, EBU-1 4307 Bionic Human-Machine Interaction Helen Meng, The Chinese University of Hong Kong Bionics is a vastly interdisciplinary enterprise that draws inspirations from nature for engineering designs and technology development. Illustrative examples abound, ranging from automobile design, material science, to impressions from science fiction. This talk presents a characterization of bionic human-machine interfaces, in terms of (1) mimicking human forms; (2) supporting human-like functions and (3) augmenting human physical and cognitive abilities...more > | ||||||||
 |
Wednesday, 08.16.06, 11 - 12, Calit2: Synthesis Center On universal classification with limited memory: How to learn the most out of a training sequence? Jacob Ziv, Technion–Israel Institute of Technology Traditionally, the analysis of information processing systems is based on certain modeling of the mechanism that generates the observed data (e.g. being a realization of some ergodic process). Based on this a-priory model, a processor (e.g. a classifier, etc.), is then optimally designed. In practice, there are many cases where not enough a-priori information about this generating model is available and one must base the design of a classifier on the observed training data only, under some complexity constraints that the classifier must comply with.... | ||||||||
 |
Friday, 08.04.06, 11 - 12, EBU-1 Rm 4307 Golden Space-Time Trellis Coded Modulation Emanuele Viterbo, Politecnico di Torino, Italy In this talk we present a trellis coded modulation scheme based on the Golden code. Set partitioning with increasing minimum determinant and coset coding of the Golden code are designed to optimize the.... | ||||||||
 |
Thursday, 06.15.06, 2 - 3, EBU1-6504 On coded and uncoded transmissions of Gaussian sources over Gaussian channels Amos Lapidoth, ETH, Zurich We consider the problem of transmitting a Gaussian source over an additive white Gaussian noise channel and the problem's extension to the transmission of correlated Gaussian sources over a multiple-access channel. We focus on the mean squared-error criterion. For the former problem the classical approaches are based on the source-channel separation theorem or on Goblick's surprisingly simple uncoded transmission scheme...more > | ||||||||
 |
Wednesday, 05.31.06, 11:00 - 12:00, EBU-1 4307 Market-based Approaches for Efficient Allocation of Network Resources Rahul Jain, UC Berkeley Many systems are characterized by complex (and often strategic) interactions between subsystems. Such systems occur in communication networks, power networks, wireless and sensor networks, etc. The strategic interactions between such subsystems often involve economic issues. This necessitates market-based algorithms for distributed control and optimization...more > | ||||||||
 |
Tuesday, 05.30.06, 2:00 - 3:00, EBU1 4307 CDMA Reverse Link Interference Cancellation Jilei Hou, Qualcomm This talk gives the principles and practice of how interference cancellation can be implemented on the CDMA reverse link of modern cellular systems. First, some simple information theoretic analysis is given. Link level analysis and simulations determine the quality of multipath channel estimation techniques. Network level simulations over a wide range of channels confirm that interference cancellation offers significant capacity gains across users, while maintaining the system link budget and stability...more > | ||||||||
 |
Thursday, 05.18.06, 3:00 - 4:00, Calit Auditorium How to structure, analyse and apply molecular interaction Sydney Brenner, Molecular Sciences Institute, Salk Institute Sydney Brenner received the 2002 Medicine Nobel Prize for establishing the C. elegans as an experimental model organism and for linking genetic analysis to cell division and organ development and aging. His many other achievements include the discovery of Messenger RNA. He will present a broad interest talk on the interaction between molecular genomics, structural functionality, and cellular molecular communication...more > | ||||||||
 |
Wednesday, 05.17.06, 3:00 - 4:00, Calit Synthesis Room Graphs, colorings and beyond in comparative genomics Sagi Snir, UC Berkeley Comparative genomics seeks to explore characteristic patterns of a set of organisms by comparing common features of the given organisms. Computational methods are a significant part in this type of discipline. In this talk I will describe the use of colored graphs to solve two problems in comparative genomics: 1. Micro-indels are small insertion or deletion events (indels) that occur during genome evolution...more > | ||||||||
 |
Monday, 05.15.06, 11:00 - 12:00, Calit2 MPR Streams and samples: new directions for processing massive data sets Andrew McGregor, University of Pennsylvania Data streams are ubiquitous. Sometimes data is processed as a stream by necessity; when estimating the statistics of network traffic flowing past a router, it is not feasible to store all the data and run conventional algorithms.... | ||||||||
 |
Thursday, 05.04.06, 10:00 am - 11:00 am, EBU1 4307 Reliable Communication in the Presence of Side Information Tie Liu, University of Illinois at Urbana-Champaign In many network communication problems including broadcast channel and distributed source coding, there is a natural strategy of converting the problem into a point-to-point problem and a series of side-information problems: one user does its own, and the rest of users treat messages sent from previous users as side information. My research studies such communication problems in two ways...more > | ||||||||
 |
Friday, 04.14.06, 10:00 - 11:00, EBU1 - 4307 Efficient Matching for Recognition and Retrieval Kristen Grauman, MIT Local image features have emerged as a powerful way to describe images of objects and scenes. Their stability under variable image conditions is critical for success in a wide range of recognition and retrieval applications. However, comparing images represented by their collections of local features is challenging, since each set may vary in cardinality and its elements lack a meaningful ordering. Existing methods compare feature sets by searching for explicit correspondences between their elements, which is too computationally expensive in many realistic settings. | ||||||||
 |
Tuesday, 04.11.06, 10:00 - 11:00, EBU1 - 4307 Role of Feedback in Communication Young-Han Kim , Stanford University Feedback plays a pivotal role in control. Without feedback, even a small amount of error can unstabilize control systems in stochastic environments. Communication systems, in which one ``controls' another's state of knowledge, are notable exceptions. By encoding data with a forward error correcting code in long blocks, one can communicate reliably over a noisy channel without any feedback, as shown by Shannon. | ||||||||
