 Date & Time: Tuesday, July 16, 2019; 12:00 PM
Speaker: Prof. Jeff Linderoth, University of WisconsinMadison
MERL Host: Arvind Raghunathan
Research Areas: Machine Learning, Optimization
Brief  Algorithms to solve mixed integer linear programs have made incredible progress in the past 20 years. Key to these advances has been a mathematical analysis of the structure of the set of feasible solutions. We argue that a similar analysis is required in the case of mixed integer quadratic programs, like those that arise in sparse optimization in machine learning. One such analysis leads to the socalled perspective relaxation, which significantly improves solution performance on separable instances. Extensions of the perspective reformulation can lead to algorithms that are equivalent to some of the most popular, modern, sparsityinducing nonconvex regularizations in variable selection. Based on joint work with Hongbo Dong (Washington State Univ. ), Oktay Gunluk (IBM), and Kun Chen (Univ. Connecticut)

 Date & Time: Thursday, February 14, 2019; 1:30 3:00 PM
Speaker: Avishai Weiss, MERL
MERL Hosts: Stefano Di Cairano; Uroš Kalabić; Avishai Weiss
Research Areas: Control, Dynamical Systems
Brief  Avishai Weiss from MERL's Control and Dynamical Systems group will give a talk at Stanford's Aeronautics and Astronautics department titled: "LowThrust GEO Satellite Station Keeping, Attitude Control, and Momentum Management via Model Predictive Control". Electric propulsion for satellites is much more fuel efficient than conventional methods. The talk will describe MERL's solution to the satellite control problems deriving from the low thrust provided by electric propulsion.

 Date & Time: Tuesday, March 6, 2018; 12:00 PM
Speaker: Scott Wisdom, Affectiva
MERL Host: Jonathan Le Roux
Research Area: Speech & Audio
Brief  Recurrent neural networks (RNNs) are effective, datadriven models for sequential data, such as audio and speech signals. However, like many deep networks, RNNs are essentially black boxes; though they are effective, their weights and architecture are not directly interpretable by practitioners. A major component of my dissertation research is explaining the success of RNNs and constructing new RNN architectures through the process of "deep unfolding," which can construct and explain deep network architectures using an equivalence to inference in statistical models. Deep unfolding yields principled initializations for training deep networks, provides insight into their effectiveness, and assists with interpretation of what these networks learn.
In particular, I will show how RNNs with rectified linear units and residual connections are a particular deep unfolding of a sequential version of the iterative shrinkagethresholding algorithm (ISTA), a simple and classic algorithm for solving L1regularized leastsquares. This equivalence allows interpretation of stateoftheart unitary RNNs (uRNNs) as an unfolded sparse coding algorithm. I will also describe a new type of RNN architecture called deep recurrent nonnegative matrix factorization (DRNMF). DRNMF is an unfolding of a sparse NMF model of nonnegative spectrograms for audio source separation. Both of these networks outperform conventional LSTM networks while also providing interpretability for practitioners.

 Date & Time: Friday, February 2, 2018; 12:00
Speaker: Dr. David Kaeli, Northeastern University
MERL Host: Abraham Goldsmith
Research Areas: Control, Optimization, Machine Learning, Speech & Audio
Brief  GPU computing is alive and well! The GPU has allowed researchers to overcome a number of computational barriers in important problem domains. But still, there remain challenges to use a GPU to target more general purpose applications. GPUs achieve impressive speedups when compared to CPUs, since GPUs have a large number of compute cores and high memory bandwidth. Recent GPU performance is approaching 10 teraflops of single precision performance on a single device. In this talk we will discuss current trends with GPUs, including some advanced features that allow them exploit multicontext grains of parallelism. Further, we consider how GPUs can be treated as cloudbased resources, enabling a GPUenabled server to deliver HPC cloud services by leveraging virtualization and collaborative filtering. Finally, we argue for for new heterogeneous workloads and discuss the role of the Heterogeneous Systems Architecture (HSA), a standard that further supports integration of the CPU and GPU into a common framework. We present a new class of benchmarks specifically tailored to evaluate the benefits of features supported in the new HSA programming model.

 Date & Time: Wednesday, February 1, 2017; 12:0013:00
Speaker: Dr. Heiga ZEN, Google
MERL Host: Chiori Hori
Research Area: Speech & Audio
Brief  Recent progress in generative modeling has improved the naturalness of synthesized speech significantly. In this talk I will summarize these generative modelbased approaches for speech synthesis such as WaveNet, a deep generative model of raw audio waveforms. We show that WaveNets are able to generate speech which mimics any human voice and which sounds more natural than the best existing TexttoSpeech systems.
See https://deepmind.com/blog/wavenetgenerativemodelrawaudio/ for further details.

 Date & Time: Tuesday, December 13, 2016; Noon
Speaker: Yue M. Lu, John A. Paulson School of Engineering and Applied Sciences, Harvard University
MERL Host: Petros Boufounos
Research Areas: Computational Sensing, Machine Learning
Brief  In this talk, we will present a framework for analyzing, in the highdimensional limit, the exact dynamics of several stochastic optimization algorithms that arise in signal and information processing. For concreteness, we consider two prototypical problems: sparse principal component analysis and regularized linear regression (e.g. LASSO). For each case, we show that the timevarying estimates given by the algorithms will converge weakly to a deterministic "limiting process" in the highdimensional limit. Moreover, this limiting process can be characterized as the unique solution of a nonlinear PDE, and it provides exact information regarding the asymptotic performance of the algorithms. For example, performance metrics such as the MSE, the cosine similarity and the misclassification rate in sparse support recovery can all be obtained by examining the deterministic limiting process. A steadystate analysis of the nonlinear PDE also reveals interesting phase transition phenomena related to the performance of the algorithms. Although our analysis is asymptotic in nature, numerical simulations show that the theoretical predictions are accurate for moderate signal dimensions.

 Date & Time: Monday, December 12, 2016; 12:00 PM
Speaker: Yanlai Chen, Department of Mathematics at the University of Massachusetts Dartmouth
Research Areas: Control, Dynamical Systems
Brief  Models of reduced computational complexity is indispensable in scenarios where a large number of numerical solutions to a parametrized problem are desired in a fast/realtime fashion. These include simulationbased design, parameter optimization, optimal control, multimodel/scale analysis, uncertainty quantification. Thanks to an offlineonline procedure and the recognition that the parameterinduced solution manifolds can be well approximated by finitedimensional spaces, reduced basis method (RBM) and reduced collocation method (RCM) can improve efficiency by several orders of magnitudes. The accuracy of the RBM solution is maintained through a rigorous a posteriori error estimator whose efficient development is critical and involves fast eigensolves.
In this talk, I will give a brief introduction of the RBM/RCM, and explain how they can be used for data compression, face recognition, and significantly delaying the curse of dimensionality for uncertainty quantification.

 Date & Time: Friday, December 2, 2016; 11:00 AM
Speaker: Prof. Waheed Bajwa, Rutgers University
MERL Host: Petros Boufounos
Research Area: Computational Sensing
Brief  While distributed information processing has a rich history, relatively less attention has been paid to the problem of collaborative learning of nonlinear geometric structures underlying data distributed across sites that are connected to each other in an arbitrary topology. In this talk, we discuss this problem in the context of collaborative dictionary learning from big, distributed data. It is assumed that a number of geographicallydistributed, interconnected sites have massive local data and they are interested in collaboratively learning a lowdimensional geometric structure underlying these data. In contrast to some of the previous works on subspacebased data representations, we focus on the geometric structure of a union of subspaces (UoS). In this regard, we propose a distributed algorithm, termed cloud KSVD, for collaborative learning of a UoS structure underlying distributed data of interest. The goal of cloud KSVD is to learn an overcomplete dictionary at each individual site such that every sample in the distributed data can be represented through a small number of atoms of the learned dictionary. Cloud KSVD accomplishes this goal without requiring communication of individual data samples between different sites. In this talk, we also theoretically characterize deviations of the dictionaries learned at individual sites by cloud KSVD from a centralized solution. Finally, we numerically illustrate the efficacy of cloud KSVD in the context of supervised training of nonlinear classsifiers from distributed, labaled training data.

 Date & Time: Friday, September 23, 2016; 12:00 PM 1:00 PM
Speaker: Dr. Earl McCune, Eridan Communications
MERL Host: Rui Ma
Research Areas: Communications, Signal Processing
Brief  To maximize the operating energy efficiency of any wireless communication link requires a global optimization not only across the entire block diagram, but also including the selected signal modulation and aspects of the link operating protocol. Achieving this global optimization is first examined for the transmitter, receiver, and baseband circuitry. Then the important aspects of signal modulation necessary to access these circuit optimizations, with examples, are presented, followed by the correspondingly important protocol aspects needed. A metric called modulationavailable energy efficiency (MAEE) compares proposed signals for compatibility with high energy efficiency objectives.

 Date & Time: Wednesday, August 17, 2016; 1 PM
Speaker: Gilles Zerah, Centre Francais en Calcul Atomique et MoleculaireIledeFrance (CFCAMIdF)
Research Areas: Applied Physics, Electronic and Photonic Devices
Brief  The first part of the talk is a highlevel review of modern technologies for atomiclevel modelling of materials. The second part discusses band gap calculations and MERL results for semiconductors.

 Date & Time: Wednesday, July 13, 2016; 2:30 PM  3:30
Speaker: Richard Lehoucq, Sandia National Laboratories
Research Areas: Computer Vision, Digital Video, Machine Learning
Brief  My presentation considers the research question of whether existing algorithms and software for the largescale sparse eigenvalue problem can be applied to problems in spectral graph theory. I first provide an introduction to several problems involving spectral graph theory. I then provide a review of several different algorithms for the largescale eigenvalue problem and briefly introduce the Anasazi package of eigensolvers.

 Date & Time: Thursday, July 7, 2016; 2:00 PM
Speaker: Dr. Sonja Glavaski, Program Director, ARPAE
MERL Host: Arvind Raghunathan
Research Area: Electric Systems
Brief  The evolution of the grid faces significant challenges if it is to integrate and accept more energy from renewable generation and other Distributed Energy Resources (DERs). To maintain grid's reliability and turn intermittent power sources into major contributors to the U.S. energy mix, we have to think about the grid differently and design it to be smarter and more flexible.
ARPAE is interested in disruptive technologies that enable increased integration of DERs by realtime adaptation while maintaining grid reliability and reducing cost for customers with smart technologies. The potential impact is significant, with projected annual energy savings of more than 3 quadrillion BTU and annual CO2 emissions reductions of more than 250 million metric tons.
This talk will identify opportunities in developing next generation control technologies and grid operation paradigms that address these challenges and enable secure, stable, and reliable transmission and distribution of electrical power. Summary of newly announced ARPAE NODES (Network Optimized Distributed Energy Systems) Program funding development of these technologies will be presented.

 Date & Time: Friday, June 3, 2016; 1:30PM  3:00PM
Speaker: Nobuaki Minematsu and Daisuke Saito, The University of Tokyo
Research Area: Speech & Audio
Brief  Speech signals covey various kinds of information, which are grouped into two kinds, linguistic and extralinguistic information. Many speech applications, however, focus on only a single aspect of speech. For example, speech recognizers try to extract only word identity from speech and speaker recognizers extract only speaker identity. Here, irrelevant features are often treated as hidden or latent by applying the probability theory to a large number of samples or the irrelevant features are normalized to have quasistandard values. In speech analysis, however, phases are usually removed, not hidden or normalized, and pitch harmonics are also removed, not hidden or normalized. The resulting speech spectrum still contains both linguistic information and extralinguistic information. Is there any good method to remove extralinguistic information from the spectrum? In this talk, our answer to that question is introduced, called speech structure. Extralinguistic variation can be modeled as feature space transformation and our speech structure is based on the transforminvariance of fdivergence. This proposal was inspired by findings in classical studies of structural phonology and recent studies of developmental psychology. Speech structure has been applied to accent clustering, speech recognition, and language identification. These applications are also explained in the talk.

 Date & Time: Friday, May 13, 2016; 12:00 PM
Speaker: Oleg Iliev, Fraunhofer Institute for Industrial Mathematics, ITWM
Research Area: Dynamical Systems
Brief  Liion batteries are widely used in automotive industry, in electronic devices, etc. In this talk we will discuss challenges related to the multiscale nature of batteries, mainly the understanding of processes in the porous electrodes at pore scale and at macroscale. A software tool for simulation of isothermal and nonisothermal electrochemical processes in porous electrodes will be presented. The pore scale simulations are done on 3D images of porous electrodes, or on computer generated 3D microstructures, which have the same characterization as real porous electrodes. Finite Volume and Finite Element algorithms for the highly nonlinear problems describing processes at pore level will be shortly presented. Model order reduction, MOR, empirical interpolation method, EIMMOR algorithms for acceleration of the computations will be discussed, as well as the reduced basis method for studying parameters dependent problems. Next, homogenization of the equations describing the electrochemical processes at the pore scale will be presented, and the results will be compared to the engineering approach based on Newman's 1D+1D model. Simulations at battery cell level will also be addressed. Finally, the challenges in modeling and simulation of degradation processes in the battery will be discussed and our first simulation results in this area will be presented.
This is joint work with A.Latz (DLR), M.Taralov, V.Taralova, J.Zausch, S.Zhang from Fraunhofer ITWM, Y.Maday from LJLL, Paris 6 and Y.Efendiev from Texas A&M.

 Date & Time: Friday, April 29, 2016; 12:00 PM  1:00 PM
Speaker: Yu Zhang, MIT
Research Area: Speech & Audio
Brief  A recurrent neural network (RNN) is a class of neural network models where connections between its neurons form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Recently the RNNbased acoustic models greatly improved automatic speech recognition (ASR) accuracy on many tasks, such as an advanced version of the RNN, which exploits a structure called longshortterm memory (LSTM). However, ASR performance with distant microphones, low resources, noisy, reverberant conditions, and on multitalker speech are still far from satisfactory as compared to humans. To address these issues, we develop new strucute of RNNs inspired by two principles: (1) the structure follows the intuition of human speech recognition; (2) the structure is easy to optimize. The talk will go beyond basic RNNs, introduce predictionadaptationcorrection RNNs (PACRNNs) and highway LSTMs (HLSTMs). It studies both unidirectional and bidirecitonal RNNs and discriminative training also applied on top the RNNs. For efficient training of such RNNs, the talk will describe two algorithms for learning their parameters in some detail: (1) LatencyControlled bidirectional model training; and (2) Two pass forward computation for sequence training. Finally, this talk will analyze the advantages and disadvantages of different variants and propose future directions.

 Date & Time: Tuesday, March 15, 2016; 12:45 PM  1:30 PM
Speaker: Prof. Hirofumi Aoki, Nagoya University
Research Area: Speech & Audio
Brief  Driving requires a complex skill that is involved with the vehicle itself (e.g., speed control and instrument operation), other road users (e.g., other vehicles, pedestrians), surrounding environment, and so on. During driving, visual cues are the main source to supply information to the brain. In order to stabilize the visual information when you are moving, the eyes move to the opposite direction based on the input to the vestibular system. This involuntary eye movement is called as the vestibuloocular reflex (VOR) and the physiological models have been studied so far. Obinata et al. found that the VOR can be used to estimate mental workload. Since then, our research group has been developing methods to quantitatively estimate mental workload during driving by means of reflex eye movement. In this talk, I will explain the basic mechanism of the reflex eye movement and how to apply for mental workload estimation. I also introduce the latest work to combine the VOR and OKR (optokinetic reflex) models for naturalistic driving environment.

 Date & Time: Tuesday, March 15, 2016; 12:00 PM  12:45 PM
Speaker: Prof. Kazuya Takeda, Nagoya University
Research Area: Speech & Audio
Brief  Thanks to advanced "internet of things" (IoT) technologies, situationspecific human behavior has become an area of development for practical applications involving signal processing. One important area of development of such practical applications is driving behavior research. Since 1999, I have been collecting driving behavior data in a wide range of signal modalities, including speech/sound, video, physical/physiological sensors, CAN bus, LIDAR and GNSS. The objective of this data collection is to evaluate how well signal models can represent human behavior while driving. In this talk, I would like to summarize our 10 years of study of driving behavior signal processing, which has been based on these signal corpora. In particular, statistical signal models of interactions between traffic contexts and driving behavior, i.e., stochastic driver modeling, will be discussed, in the context of risky lane change detection. I greatly look forward to discussing the scalability of such corpusbased approaches, which could be applied to almost any traffic situation.

 Date & Time: Tuesday, February 16, 2016; 12:00 PM  1:00 PM
Speaker: Dr. Najim Dehak, MIT
Research Area: Speech & Audio
Brief  Recently, there has been a great increase of interest in the field of emotion recognition based on different human modalities, such as speech, heart rate etc. Emotion recognition systems can be very useful in several areas, such as medical and telecommunications. In the medical field, identifying the emotions can be an important tool for detecting and monitoring patients with mental health disorder. In addition, the identification of the emotional state from voice provides opportunities for the development of automated dialogue system capable of producing reports to the physician based on frequent phone communication between the system and the patients. In this talk, we will describe a health related application of using emotion recognition system based on human voices in order to detect and monitor the emotion state of people.

 Date & Time: Monday, November 23, 2015; 12:00 PM
Speaker: Manuchehr Aminian, University of North Carolina, Chapel Hill Brief  The classic work by G.I. Taylor describes the enhanced longitudinal diffusivity of a passive tracer subjected to laminar pipe flow. Much work since then has gone into extending this result particularly in calculating the evolution of the scalar variance. However, less work has been done to describe the evolution of asymmetry in the distribution. We present the results from a modeling effort to understand how the higher moments of the tracer distribution depend on geometry based off of explicit results in the circular pipe. We do this via analysis of "channellimiting" geometries (rectangular ducts and elliptical pipes parameterized by their aspect ratio), using both new analytical tools and Monte Carlo simulation, which have revealed a wealth of nontrivial behavior of the distributions at short and intermediate time.

 Date & Time: Friday, October 18, 2013; 12:00 PM
Speaker: Dr. Shreyas Sundaram, University of Waterloo
MERL Host: Mouhacine Benosman Brief  This talk will describe a method to stabilize a plant with a network of resourceconstrained wireless nodes. As opposed to traditional networked control schemes where the nodes simply route information to and from a dedicated controller, our approach treats the network itself as the controller. Specifically, we formulate a strategy where each node repeatedly updates its state to be a linear combination of the states of neighboring nodes. We show that this causes the entire network to behave as a linear dynamical system, with sparsity constraints imposed by the network topology. We provide a numerical design procedure to determine the appropriate linear combinations for each node so that the transmissions of the nodes closest to the actuators are stabilizing. We also make connections to decentralized control theory and the concept of fixed modes to provide topological conditions under which stabilization is possible. We show that this "Wireless Control Network" requires low computational and communication overhead, simplifies transmission scheduling, and enables compositional design. We also consider the issue of security in this control scheme. Using structured system theory, we show that a certain number of malicious or misbehaving nodes can be detected and identified provided that the connectivity of the network is sufficiently high.
