- Date & Time: Thursday, December 9, 2021; 1:00pm - 5:30pm EST
Location: Virtual Event
Speaker: Prof. Melanie Zeilinger, ETH
Research Areas: Applied Physics, Artificial Intelligence, Communications, Computational Sensing, Computer Vision, Control, Data Analytics, Dynamical Systems, Electric Systems, Electronic and Photonic Devices, Machine Learning, Multi-Physical Modeling, Optimization, Robotics, Signal Processing, Speech & Audio, Digital Video, Human-Computer Interaction, Information Security
Brief - MERL is excited to announce the second keynote speaker for our Virtual Open House 2021:
Prof. Melanie Zeilinger from ETH .
Our virtual open house will take place on December 9, 2021, 1:00pm - 5:30pm (EST).
Join us to learn more about who we are, what we do, and discuss our internship and employment opportunities. Prof. Zeilinger's talk is scheduled for 3:15pm - 3:45pm (EST).
Registration: https://mailchi.mp/merl/merlvoh2021
Keynote Title: Control Meets Learning - On Performance, Safety and User Interaction
Abstract: With increasing sensing and communication capabilities, physical systems today are becoming one of the largest generators of data, making learning a central component of autonomous control systems. While this paradigm shift offers tremendous opportunities to address new levels of system complexity, variability and user interaction, it also raises fundamental questions of learning in a closed-loop dynamical control system. In this talk, I will present some of our recent results showing how even safety-critical systems can leverage the potential of data. I will first briefly present concepts for using learning for automatic controller design and for a new safety framework that can equip any learning-based controller with safety guarantees. The second part will then discuss how expert and user information can be utilized to optimize system performance, where I will particularly highlight an approach developed together with MERL for personalizing the motion planning in autonomous driving to the individual driving style of a passenger.
-
- Date & Time: Thursday, December 9, 2021; 1:00pm - 5:30pm EST
Location: Virtual Event
Speaker: Prof. Ashok Veeraraghavan, Rice University
Research Areas: Applied Physics, Artificial Intelligence, Communications, Computational Sensing, Computer Vision, Control, Data Analytics, Dynamical Systems, Electric Systems, Electronic and Photonic Devices, Machine Learning, Multi-Physical Modeling, Optimization, Robotics, Signal Processing, Speech & Audio, Digital Video, Human-Computer Interaction, Information Security
Brief - MERL is excited to announce the first keynote speaker for our Virtual Open House 2021:
Prof. Ashok Veeraraghavan from Rice University.
Our virtual open house will take place on December 9, 2021, 1:00pm - 5:30pm (EST).
Join us to learn more about who we are, what we do, and discuss our internship and employment opportunities. Prof. Veeraraghavan's talk is scheduled for 1:15pm - 1:45pm (EST).
Registration: https://mailchi.mp/merl/merlvoh2021
Keynote Title: Computational Imaging: Beyond the limits imposed by lenses.
Abstract: The lens has long been a central element of cameras, since its early use in the mid-nineteenth century by Niepce, Talbot, and Daguerre. The role of the lens, from the Daguerrotype to modern digital cameras, is to refract light to achieve a one-to-one mapping between a point in the scene and a point on the sensor. This effect enables the sensor to compute a particular two-dimensional (2D) integral of the incident 4D light-field. We propose a radical departure from this practice and the many limitations it imposes. In the talk we focus on two inter-related research projects that attempt to go beyond lens-based imaging.
First, we discuss our lab’s recent efforts to build flat, extremely thin imaging devices by replacing the lens in a conventional camera with an amplitude mask and computational reconstruction algorithms. These lensless cameras, called FlatCams can be less than a millimeter in thickness and enable applications where size, weight, thickness or cost are the driving factors. Second, we discuss high-resolution, long-distance imaging using Fourier Ptychography, where the need for a large aperture aberration corrected lens is replaced by a camera array and associated phase retrieval algorithms resulting again in order of magnitude reductions in size, weight and cost. Finally, I will spend a few minutes discussing how the wholistic computational imaging approach can be used to create ultra-high-resolution wavefront sensors.
-
- Date & Time: Tuesday, November 16, 2021; 11:00 AM EST
Speaker: Thomas Schön, Uppsala University
MERL Host: Karl Berntorp
Research Areas: Dynamical Systems, Machine Learning
Abstract 
While deep learning-based classification is generally addressed using standardized approaches, this is really not the case when it comes to the study of regression problems. There are currently several different approaches used for regression and there is still room for innovation. We have developed a general deep regression method with a clear probabilistic interpretation. The basic building block in our construction is an energy-based model of the conditional output density p(y|x), where we use a deep neural network to predict the un-normalized density from input-output pairs (x, y). Such a construction is also commonly referred to as an implicit representation. The resulting learning problem is challenging and we offer some insights on how to deal with it. We show good performance on several computer vision regression tasks, system identification problems and 3D object detection using laser data.
-
- Date & Time: Thursday, December 9, 2021; 100pm-5:30pm (EST)
Location: Virtual Event
Research Areas: Applied Physics, Artificial Intelligence, Communications, Computational Sensing, Computer Vision, Control, Data Analytics, Dynamical Systems, Electric Systems, Electronic and Photonic Devices, Machine Learning, Multi-Physical Modeling, Optimization, Robotics, Signal Processing, Speech & Audio, Digital Video, Human-Computer Interaction, Information Security
Brief - Mitsubishi Electric Research Laboratories cordially invites you to join our Virtual Open House, on December 9, 2021, 1:00pm - 5:30pm (EST).
The event will feature keynotes, live sessions, research area booths, and time for open interactions with our researchers. Join us to learn more about who we are, what we do, and discuss our internship and employment opportunities.
Registration: https://mailchi.mp/merl/merlvoh2021
-
- Date & Time: Tuesday, November 9, 2021; 1:00 PM EST
Speaker: Prof. Marco Di Renzo, CNRS & Paris-Saclay University
Research Areas: Communications, Electronic and Photonic Devices, Signal Processing
Abstract 
A Reconfigurable Intelligent Surface (RIS) is a planar structure that is engineered to have properties that enable the dynamic control of the electromagnetic waves. In wireless communications and networks, RISs are an emerging technology for realizing programmable and reconfigurable wireless propagation environments through nearly passive and tunable signal transformations. RIS-assisted programmable wireless environments are a multidisciplinary research endeavor. This presentation is aimed to report the latest research advances on modeling, analyzing, and optimizing RISs for wireless communications with focus on electromagnetically consistent models, analytical frameworks, and optimization algorithms.
-
- Date & Time: Tuesday, November 2, 2021; 1:00 PM EST
Speaker: Dr. Hsiao-Yu (Fish) Tung, MIT BCS
Research Areas: Artificial Intelligence, Computer Vision, Machine Learning, Robotics
Abstract 
Current state-of-the-art CNNs can localize and name objects in internet photos, yet, they miss the basic knowledge that a two-year-old toddler has possessed: objects persist over time despite changes in the observer’s viewpoint or during cross-object occlusions; objects have 3D extent; solid objects do not pass through each other. In this talk, I will introduce neural architectures that learn to parse video streams of a static scene into world-centric 3D feature maps by disentangling camera motion from scene appearance. I will show the proposed architectures learn object permanence, can imagine RGB views from novel viewpoints in truly novel scenes, can conduct basic spatial reasoning and planning, can infer affordability in sentences, and can learn geometry-aware 3D concepts that allow pose-aware object recognition to happen with weak/sparse labels. Our experiments suggest that the proposed architectures are essential for the models to generalize across objects and locations, and it overcomes many limitations of 2D CNNs. I will show how we can use the proposed 3D representations to build machine perception and physical understanding more close to humans.
-
- Date & Time: Tuesday, October 12, 2021; 1:00 PM EST
Speaker: Prof. Greg Ongie, Marquette University
MERL Host: Hassan Mansour
Research Areas: Computational Sensing, Machine Learning, Signal Processing
Abstract 
Deep learning is emerging as powerful tool to solve challenging inverse problems in computational imaging, including basic image restoration tasks like denoising and deblurring, as well as image reconstruction problems in medical imaging. This talk will give an overview of the state-of-the-art supervised learning techniques in this area and discuss two recent innovations: deep equilibrium architectures, which allows one to train an effectively infinite-depth reconstruction network; and model adaptation methods, that allow one to adapt a pre-trained reconstruction network to changes in the imaging forward model at test time.
-
- Date & Time: Tuesday, September 28, 2021; 1:00 PM EST
Speaker: Dr. Ruohan Gao, Stanford University
MERL Host: Gordon Wichern
Research Areas: Computer Vision, Machine Learning, Speech & Audio
Abstract 
While computer vision has made significant progress by "looking" — detecting objects, actions, or people based on their appearance — it often does not listen. Yet cognitive science tells us that perception develops by making use of all our senses without intensive supervision. Towards this goal, in this talk I will present my research on audio-visual learning — We disentangle object sounds from unlabeled video, use audio as an efficient preview for action recognition in untrimmed video, decode the monaural soundtrack into its binaural counterpart by injecting visual spatial information, and use echoes to interact with the environment for spatial image representation learning. Together, these are steps towards multimodal understanding of the visual world, where audio serves as both the semantic and spatial signals. In the end, I will also briefly talk about our latest work on multisensory learning for robotics.
-
- Date & Time: Tuesday, September 14, 2021; 1:00 PM EST
Speaker: Prof. David Bergman, University of Connecticut
MERL Host: Arvind Raghunathan
Research Areas: Data Analytics, Machine Learning, Optimization
Abstract 
The integration of machine learning and optimization opens the door to new modeling paradigms that have already proven successful across a broad range of industries. Sports betting is a particularly exciting application area, where recent advances in both analytics and optimization can provide a lucrative edge. In this talk we will discuss three algorithmic sports betting games where combinations of machine learning and optimization have netted me significant winnings.
-
- Date & Time: Tuesday, February 16, 2021; 11:00-12:00
Speaker: Prof. Pere Gilabert, Universitat Politecnica de Catalunya, Barcelona, Spain
Research Areas: Communications, Electronic and Photonic Devices, Machine Learning, Signal Processing
Abstract - Digital predistortion (DPD) linearization is the most common and spread solution to cope with power amplifiers (PA) inherent linearity versus efficiency trade-off. The use of new radio 5G spectrally efficient signals with high peak-to-average power ratios (PAPR) occupying wider bandwidths only aggravates such compromise. When considering wide bandwidth signals, carrier aggregation or multi-band configurations in high efficient transmitter architectures, such as Doherty PAs, load-modulated balanced amplifiers, envelope tracking PAs or outphasing transmitters, the number of parameters required in the DPD model to compensate for both nonlinearities and memory effects can be unacceptably high. This has a negative impact in the DPD model extraction/adaptation, because it increases the computational complexity and drives to over-fitting and uncertainty.
This talk will discuss the use of machine learning techniques for DPD linearization. The use of artificial neural networks (ANNs) for adaptive DPD linearization and approaches to reduce the coefficients adaptation time will be discussed. In addition, an overview on several feature-extraction techniques used to reduce the number of parameters of the DPD linearization system as well as to ensure proper, well-conditioned estimation for related variables will be presented.
-
- Date & Time: Wednesday, December 9, 2020; 1:00-5:00PM EST
Location: Virtual
MERL Contacts: Elizabeth Phillips; Anthony Vetro
Research Areas: Applied Physics, Artificial Intelligence, Communications, Computational Sensing, Computer Vision, Control, Data Analytics, Dynamical Systems, Electric Systems, Electronic and Photonic Devices, Machine Learning, Multi-Physical Modeling, Optimization, Robotics, Signal Processing, Speech & Audio
-
- Date & Time: Tuesday, August 25, 2020; 11:00 AM
Speaker: Prof. James Hwang, Cornell University
Research Areas: Applied Physics, Electronic and Photonic Devices
Abstract 
Microwave is not just for cooking, smart cars, or mobile phones. We can take advantage of the wide electromagnetic spectrum to do wonderful things that are more vital to our lives. For example, microwave ablation of cancer tumor is already in wide use, and microwave remote monitoring of vital signs is becoming more important as the population ages. This talk will focus on a biomedical use of microwave at the single-cell level. At low power, microwave can readily penetrate a cell membrane to interrogate what is inside a cell, without cooking it or otherwise hurting it. It is currently the fastest, most compact, and least costly way to tell whether a cell is alive or dead. On the other hand, at higher power but lower frequency, the electromagnetic signal can interact strongly with the cell membrane to drill temporary holes of nanometer size. The nanopores allow drugs to diffuse into the cell and, based on the reaction of the cell, individualized medicine can be developed and drug development can be sped up in general. Conversely, the nanopores allow strands of DNA molecules to be pulled out of the cell without killing it, which can speed up genetic engineering. Lastly, by changing both the power and frequency of the signal, we can have either positive or negative dielectrophoresis effects, which we have used to coerce a live cell to the examination table of Dr. Microwave, then usher it out after examination. These interesting uses of microwave and the resulted fundamental knowledge about biological cells will be explored in the talk.
-
- Date & Time: Tuesday, July 14, 2020; 11:00 AM
Speaker: Hanrui Wang, MIT
Research Areas: Electronic and Photonic Devices, Machine Learning
Abstract - Automatic transistor sizing is a challenging problem in circuit design due to the large design space, complex performance trade-offs, and fast technological advancements. Although there has been plenty of work on transistor sizing targeting on one circuit, limited research has been done on transferring the knowledge from one circuit to another to reduce the re-design overhead. In this work, we present GCN-RL Circuit Designer, leveraging reinforcement learning (RL) to transfer the knowledge between different technology nodes and topologies. Moreover, inspired by the simple fact that circuit is a graph, we learn on the circuit topology representation with graph convolutional neural networks (GCN). The GCN-RL agent extracts features of the topology graph whose vertices are transistors, edges are wires. Our learning-based optimization consistently achieves the highest Figures of Merit (FoM) on four different circuits compared with conventional black-box optimization methods (Bayesian Optimization, Evolutionary Algorithms), random search, and human expert designs. Experiments on transfer learning between five technology nodes and two circuit topologies demonstrate that RL with transfer learning can achieve much higher FoMs than methods without knowledge transfer. Our transferable optimization method makes transistor sizing and design porting more effective and efficient. The work is accepted to DAC 2020.
-
- Date & Time: Thursday, May 7, 2020; 12:00 PM
Speaker: Christopher Rackauckas, MIT
MERL Host: Christopher R. Laughman
Research Areas: Machine Learning, Multi-Physical Modeling, Optimization
Abstract - In the context of science, the well-known adage "a picture is worth a thousand words" might well be "a model is worth a thousand datasets." Scientific models, such as Newtonian physics or biological gene regulatory networks, are human-driven simplifications of complex phenomena that serve as surrogates for the countless experiments that validated the models. Recently, machine learning has been able to overcome the inaccuracies of approximate modeling by directly learning the entire set of nonlinear interactions from data. However, without any predetermined structure from the scientific basis behind the problem, machine learning approaches are flexible but data-expensive, requiring large databases of homogeneous labeled training data. A central challenge is reco nciling data that is at odds with simplified models without requiring "big data". In this talk we discuss a new methodology, universal differential equations (UDEs), which augment scientific models with machine-learnable structures for scientifically-based learning. We show how UDEs can be utilized to discover previously unknown governing equations, accurately extrapolate beyond the original data, and accelerate model simulation, all in a time and data-efficient manner. This advance is coupled with open-source software that allows for training UDEs which incorporate physical constraints, delayed interactions, implicitly-defined events, and intrinsic stochasticity in the model. Our examples show how a diverse set of computationally-difficult modeling issues across scientific disciplines, from automatically discovering biological mechanisms to accelerating climate simulations by 15,000x, can be handled by training UDEs.
-
- Date & Time: Thursday, May 7, 2020; 11:00 AM
Speaker: Prof. Petar Popovski, Aalborg University, Denmark
MERL Host: Toshiaki Koike-Akino
Research Areas: Artificial Intelligence, Communications, Machine Learning, Signal Processing, Information Security
Abstract 
The wireless landscape evolves towards supporting a large population of connections for humans and machines with very diverse features and requirements. Perhaps the main motivation of 5G wireless systems is its flexibility to support heterogeneous connectivity requirements: enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). However, this classification is rather limited and is currently undergoing a revision within the research community. The first part of this talk will discuss how this heterogeneity can be revised and which opportunities it opens with respect to spectrum usage. The second part of the talk will deal with performance guarantees of wireless services and, specifically, ultra-reliable communication and outline the importance of machine learning in that context. The final part of the talk will provide a broader view on the evolution of wireless connectivity, including aspects that are implied by the resistance to the deployment of 5G, but also the new opportunities that can transform the way we build and utilize connected systems.
-
- Date & Time: Tuesday, July 16, 2019; 12:00 PM
Speaker: Prof. Jeff Linderoth, University of Wisconsin-Madison
MERL Host: Arvind Raghunathan
Research Areas: Machine Learning, Optimization
Abstract 
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 so-called 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, sparsity-inducing non-convex regularizations in variable selection. Based on joint work with Hongbo Dong (Washington State Univ. ), Oktay Gunluk (IBM), and Kun Chen (Univ. Connecticut).
-
- Date & Time: Tuesday, June 18, 2019; 12:00PM
Location: 201 Broadway, Cambridge, MA
Speaker: Beverly Shultz
MERL Contact: Elizabeth Phillips Brief - MERL hosted its annual "Women In Science Luncheon" to celebrate and inspire the Lab's team of female researchers, PhD student interns and members of the HQ staff. Beverly Shultz, author of "Skip the Typing Test, I’ll Manage the Software-A Woman’s Pioneering Journey in High Tech” joined the event to share her insights as a successful female engineer, who brought passion and technology business acumen to the male-dominated computer revolution.
Beverly was a former Vice President of Engineering at Mitsubishi Electric of America and responsible to produce several versions of an early volume rendering product. She was the first female recipient of the MELCO’s President’s Award for Technology, for this work.
-
- Date & Time: Thursday, February 14, 2019; 1:30 -3:00 PM
Speaker: Avishai Weiss, MERL
MERL Hosts: Stefano Di Cairano; Avishai Weiss
Research Area: Control
Abstract - Avishai Weiss from MERL's Control and Dynamical Systems group will give a talk at Stanford's Aeronautics and Astronautics department titled: "Low-Thrust 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: Wednesday, September 26, 2018 - Friday, September 28, 2018
Location: Houston, Texas
MERL Contacts: Chiori Hori; Elizabeth Phillips
Research Areas: Artificial Intelligence, Computer Vision, Machine Learning
Brief - "MERL, in partnership with Mitsubishi Electric was a Gold Sponsor of the Grace Hopper Celebration 2018 (GHC18) held in Houston, TX on September 26-28th. Presented by AnitaB.org and the Association for Computing Machinery, this is world's largest gathering of women technologists. Chiori Hori and Elizabeth Phillips from MERL, and Yoshiyuki Umei, Jared Baker and Lien Randle from MEUS, proudly represented Mitsubishi Electric at the recruiting expo, that drew over 20,000 female technologists this year.
-
- Date: Sunday, October 28, 2018 - Friday, November 2, 2018
Location: Banff International Research Station (BIRS), Alberta, Canada
MERL Contact: Petros T. Boufounos
Research Areas: Computational Sensing, Signal Processing
Brief - Dr. Petros Boufounos, Prof. Stark Draper (U. of Toronto) and Prof. Yonina Eldar (Technion) are co-organizing a workshop on the intersection of Information Theory and Signal Processing. The 5-day workshop will take place Oct. 28 - Nov. 2 at the Banff International Research Station (BIRS) in Alberta, Canada. The workshop schedule includes invited talks from prominent researchers in the two fields, coming together from all over the world. Parts of the workshop will be streamed live through the BIRS website.
-
- Date & Time: Thursday, November 29, 2018; 4-6pm
Location: 201 Broadway, 8th floor, Cambridge, MA
MERL Contacts: Elizabeth Phillips; Anthony Vetro
Research Areas: Applied Physics, Artificial Intelligence, Communications, Computational Sensing, Computer Vision, Control, Data Analytics, Dynamical Systems, Electric Systems, Electronic and Photonic Devices, Machine Learning, Multi-Physical Modeling, Optimization, Robotics, Signal Processing, Speech & Audio
Brief - Snacks, demos, science: On Thursday 11/29, Mitsubishi Electric Research Labs (MERL) will host an open house for graduate+ students interested in internships, post-docs, and research scientist positions. The event will be held from 4-6pm and will feature demos & short presentations in our main areas of research including artificial intelligence, robotics, computer vision, speech processing, optimization, machine learning, data analytics, signal processing, communications, sensing, control and dynamical systems, as well as multi-physyical modeling and electronic devices. MERL is a high impact publication-oriented research lab with very extensive internship and university collaboration programs. Most internships lead to publication; many of our interns and staff have gone on to notable careers at MERL and in academia. Come mix with our researchers, see our state of the art technologies, and learn about our research opportunities. Dress code: casual, with resumes.
Pre-registration for the event is strongly encouraged:
merlopenhouse.eventbrite.com
Current internship and employment openings:
www.merl.com/internship/openings
www.merl.com/employment/employment
Information about working at MERL:
www.merl.com/employment.
-
- Date & Time: Monday, October 8, 2018 - Thursday, October 11, 2018; 8am-5pm
Location: MIT Samberg Conference Center, Cambridge, MA
MERL Contact: Christopher R. Laughman
Research Areas: Control, Multi-Physical Modeling
Brief - The 2018 American Modelica Conference, the first North American conference focused on the Modelica multiphysics modeling language, will be held on Tuesday and Wednesday, October 9-10, 2018 at the Samberg Conference Center at MIT in Cambridge, MA. Chris Laughman, a team leader in the Multiphysical Systems and Devices group, is the local chair for the conference.
This conference will feature over 40 papers and user presentations on the Modelica language and its application to a wide variety of problem domains, including thermofluid, aerospace, automotive, and energy systems. There will also be 2 keynote addresses by John McKibben (Proctor & Gamble) and Hilding Elmqvist (Mogram AB). Nearly 100 attendees from 11 different countries have already registered for the conference, and it promises to be a very educational experience.
MERL is also hosting two free workshops on October 8 to provide opportunities to engineers looking to increase their familiarity with the language and its applications. An introductory workshop will be led by engineers from Modelon during that morning, and then a second workshop on the application of Modelica to building systems will be led by Michael Wetter from Lawrence Berkeley National Labs in the afternoon. MERL will also host a Modelica user meeting on October 11 that will provide more details and discussion about trends in the use and development of Modelica in the larger engineering community.
-
- Date: Thursday, October 18, 2018
Location: Google, Cambridge, MA
MERL Contact: Jonathan Le Roux
Research Area: Speech & Audio
Brief - SANE 2018, a one-day event gathering researchers and students in speech and audio from the Northeast of the American continent, will be held on Thursday October 18, 2018 at Google, in Cambridge, MA. MERL is one of the organizers and sponsors of the workshop.
It is the 7th edition in the SANE series of workshops, which started at MERL in 2012. Since the first edition, the audience has steadily grown, with a record 180 participants in 2017.
SANE 2018 will feature invited talks by leading researchers from the Northeast, as well as from the international community. It will also feature a lively poster session, open to both students and researchers.
-
- Date: Thursday, July 19, 2018
Location: MERL
MERL Contacts: Elizabeth Phillips; Jinyun Zhang Brief - We hosted the 4th Annual "Women in Science at MERL," event on July 19th. This year we celebrated the contributions of the eleven female interns, three female researchers, and some female members of HQ staff. MERL executives, managers and researchers participated in the event. MERL's interns and researchers were asked probing questions about how they are fulfilled in their work and how they facilitate innovation. This resulted in every participant feeling as though they were moving their field of science forward. Everyone left feeling inspired.
-
- Date & Time: Tuesday, March 6, 2018; 12:00 PM
Speaker: Scott Wisdom, Affectiva
MERL Host: Jonathan Le Roux
Research Area: Speech & Audio
Abstract 
Recurrent neural networks (RNNs) are effective, data-driven 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 shrinkage-thresholding algorithm (ISTA), a simple and classic algorithm for solving L1-regularized least-squares. This equivalence allows interpretation of state-of-the-art 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 (DR-NMF). DR-NMF 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.
-