MERL - Mitsubishi Electric Research Laboratories

Matthew Brand

Matthew Brand
Matt's Quick Links

  • Biography

    Matt develops and analyzes optimization algorithms for problems in logistics, control, perception, data-mining, and learning. Notable results include methods for parallel solution of quadratic programs, recomposing photos by re-arranging pixels, nonlinear dimensionality reduction, online singular value decomposition, 3D shape-from-video, and learning concise models of data. In addition to academic "best paper" awards, this work has garnered several industrial awards for commercialized technologies.

  • Recent News & Events

    •  NEWS    MERL researchers presenting workshop papers at NeurIPS 2022
      Date: December 2, 2022 - December 8, 2022
      MERL Contacts: Matthew Brand; Toshiaki Koike-Akino; Jing Liu; Saviz Mowlavi; Kieran Parsons; Ye Wang
      Research Areas: Artificial Intelligence, Control, Dynamical Systems, Machine Learning, Signal Processing
      Brief
      • In addition to 5 papers in recent news (https://www.merl.com/news/news-20221129-1450), MERL researchers presented 2 papers at the NeurIPS Conference Workshop, which was held Dec. 2-8. NeurIPS is one of the most prestigious and competitive international conferences in machine learning.

        - “Optimal control of PDEs using physics-informed neural networks” by Saviz Mowlavi and Saleh Nabi

        Physics-informed neural networks (PINNs) have recently become a popular method for solving forward and inverse problems governed by partial differential equations (PDEs). By incorporating the residual of the PDE into the loss function of a neural network-based surrogate model for the unknown state, PINNs can seamlessly blend measurement data with physical constraints. Here, we extend this framework to PDE-constrained optimal control problems, for which the governing PDE is fully known and the goal is to find a control variable that minimizes a desired cost objective. We validate the performance of the PINN framework by comparing it to state-of-the-art adjoint-based optimization, which performs gradient descent on the discretized control variable while satisfying the discretized PDE.

        - “Learning with noisy labels using low-dimensional model trajectory” by Vasu Singla, Shuchin Aeron, Toshiaki Koike-Akino, Matthew E. Brand, Kieran Parsons, Ye Wang

        Noisy annotations in real-world datasets pose a challenge for training deep neural networks (DNNs), detrimentally impacting generalization performance as incorrect labels may be memorized. In this work, we probe the observations that early stopping and low-dimensional subspace learning can help address this issue. First, we show that a prior method is sensitive to the early stopping hyper-parameter. Second, we investigate the effectiveness of PCA, for approximating the optimization trajectory under noisy label information. We propose to estimate the low-rank subspace through robust and structured variants of PCA, namely Robust PCA, and Sparse PCA. We find that the subspace estimated through these variants can be less sensitive to early stopping, and can outperform PCA to achieve better test error when trained on noisy labels.

        - In addition, new MERL researcher, Jing Liu, also presented a paper entitled “CoPur: Certifiably Robust Collaborative Inference via Feature Purification" based on his previous work before joining MERL. His paper was elected as a spotlight paper to be highlighted in lightening talks and featured paper panel.
    •  
    •  TALK    [MERL Seminar Series 2022] Extreme optics design as a large-scale optimization problem
      Date & Time: Tuesday, February 8, 2022; 1:00 PM EST
      Speaker: Raphaël Pestourie, MIT
      MERL Host: Matthew Brand
      Research Areas: Applied Physics, Electronic and Photonic Devices, Optimization
      Abstract
      • Thin large-area structures with aperiodic subwavelength patterns can unleash the full power of Maxwell’s equations for focusing light and a variety of other wave transformation or optical applications. Because of their irregularity and large scale, capturing the full scattering through these devices is one of the most challenging tasks for computational design: enter extreme optics! This talk will present ways to harness the full computational power of modern large-scale optimization in order to design optical devices with thousands or millions of free parameters. We exploit various methods of domain-decomposition approximations, supercomputer-scale topology optimization, laptop-scale “surrogate” models based on Chebyshev interpolation and/or new scientific machine learning models, and other techniques to attack challenging problems: achromatic lenses that simultaneously handle many wavelengths and angles, “deep” images, hyperspectral imaging, and more.
    •  

    See All News & Events for Matt

  • Research Highlights

  • Internships with Matt

    • SA1959: Metasurfaces for machine vision

      We seek highly qualified candidates for research on co-design and optimization of metasurfaces and machine vision algorithms, with a particular interest in polarization. Strong candidates will have a background in metasurface optics, fluency with FDTD and RCWA simulation tools, and some familiarity with optimization methods used in computer vision and machine learning.

    See All Internships at MERL

  • MERL Publications

    •  Kojima, K., Koike-Akino, T., Wang, Y., Jung Minwoo, , Brand, M., "Inverse design of two-dimensional freeform metagrating using an adversarial conditional variational autoencoder", SPIE Photonics West, January 2023.
      BibTeX TR2023-004 PDF
      • @inproceedings{Kojima2023jan,
      • author = {Kojima, Keisuke and Koike-Akino, Toshiaki and Wang, Ye and Jung Minwoo and Brand, Matthew},
      • title = {Inverse design of two-dimensional freeform metagrating using an adversarial conditional variational autoencoder},
      • booktitle = {SPIE Photonics West},
      • year = 2023,
      • month = jan,
      • url = {https://www.merl.com/publications/TR2023-004}
      • }
    •  Singla, V., Aeron, S., Koike-Akino, T., Parsons, K., Brand, M., Wang, Y., "Learning with noisy labels using low-dimensional model trajectory", NeurIPS 2022 Workshop on Distribution Shifts (DistShift), December 2022.
      BibTeX TR2022-156 PDF
      • @inproceedings{Singla2022dec,
      • author = {Singla, Vasu and Aeron, Shuchin and Koike-Akino, Toshiaki and Parsons, Kieran and Brand, Matthew and Wang, Ye},
      • title = {Learning with noisy labels using low-dimensional model trajectory},
      • booktitle = {NeurIPS 2022 Workshop on Distribution Shifts (DistShift)},
      • year = 2022,
      • month = dec,
      • url = {https://www.merl.com/publications/TR2022-156}
      • }
    •  Yu, X., Smedemark-Margulies, N., Aeron, S., Koike-Akino, T., Moulin, P., Brand, M., Parsons, K., Wang, Y., "Improving Adversarial Robustness by Learning Shared Information", Pattern Recognition, DOI: 10.1016/​j.patcog.2022.109054, Vol. 134, pp. 109054, November 2022.
      BibTeX TR2022-141 PDF
      • @article{Yu2022nov,
      • author = {Yu, Xi and Smedemark-Margulies, Niklas and Aeron, Shuchin and Koike-Akino, Toshiaki and Moulin, Pierre and Brand, Matthew and Parsons, Kieran and Wang, Ye},
      • title = {Improving Adversarial Robustness by Learning Shared Information},
      • journal = {Pattern Recognition},
      • year = 2022,
      • volume = 134,
      • pages = 109054,
      • month = nov,
      • doi = {10.1016/j.patcog.2022.109054},
      • issn = {0031-3203},
      • url = {https://www.merl.com/publications/TR2022-141}
      • }
    •  Liu, Bryan, Koike-Akino, Toshiaki, Wang, Ye, Kim, Kyeong Jin, Brand, Matthew, Aeron, Shuchin, Parsons, Kieran, "Data Privacy and Protection on Deep Leakage from Gradients by Layer-Wise Pruning", Tech. Rep. TR2022-081, Mitsubishi Electric Research Laboratories, Cambridge, MA, August 2022.
      BibTeX TR2022-081 PDF
      • @techreport{MERL_TR2022-081,
      • author = {Liu, Bryan; Koike-Akino, Toshiaki; Wang, Ye; Kim, Kyeong Jin; Brand, Matthew; Aeron, Shuchin; Parsons, Kieran},
      • title = {Data Privacy and Protection on Deep Leakage from Gradients by Layer-Wise Pruning},
      • institution = {MERL - Mitsubishi Electric Research Laboratories},
      • address = {Cambridge, MA 02139},
      • number = {TR2022-081},
      • month = aug,
      • year = 2022,
      • url = {https://www.merl.com/publications/TR2022-081/}
      • }
    •  Kojima, K., Jung, M., Koike-Akino, T., Wang, Y., Brand, M., Parsons, K., "Deep Transfer Learning for Nanophotonic Device Design", Conference on Lasers and Electro-Optics (CLEO) Pasific Rim, July 2022.
      BibTeX TR2022-107 PDF
      • @inproceedings{Kojima2022jul,
      • author = {Kojima, Keisuke and Jung, Minwoo and Koike-Akino, Toshiaki and Wang, Ye and Brand, Matthew and Parsons, Kieran},
      • title = {Deep Transfer Learning for Nanophotonic Device Design},
      • booktitle = {Conference on Lasers and Electro-Optics (CLEO) Pasific Rim},
      • year = 2022,
      • month = jul,
      • url = {https://www.merl.com/publications/TR2022-107}
      • }
    See All Publications for Matt

  • Software Downloads

  • Videos

  • MERL Issued Patents

    • Title: "System and Method for Generating Optimal Lattice Tool Paths"
      Inventors: Brand, Matthew E.
      Patent No.: 11,392,105
      Issue Date: Jul 19, 2022
    • Title: "Machine Learning via Double Layer Optimization"
      Inventors: Zhang, Ziming; Brand, Matthew E.
      Patent No.: 11,170,301
      Issue Date: Nov 9, 2021
    • Title: "Uniform-irradiance extended-source freeforms"
      Inventors: Brand, Matthew E.; Birch, Daniel
      Patent No.: 10,995,932
      Issue Date: May 4, 2021
    • Title: "Methods and Systems for Freeform Irradiance Tailoring for Light Fields"
      Inventors: Brand, Matthew E.; Birch, Daniel
      Patent No.: 10,837,621
      Issue Date: Nov 17, 2020
    • Title: "Compound Optics with Freeform Optical Surface"
      Inventors: Brand, Matthew E.
      Patent No.: 10,234,689
      Issue Date: Mar 19, 2019
    • Title: "Freeform Optical Surface for Producing Sharp-Edged Irradiance Patterns"
      Inventors: Brand, Matthew E.
      Patent No.: 10,119,679
      Issue Date: Nov 6, 2018
    • Title: "Tailored Freeform Optical Surface"
      Inventors: Brand, Matthew E.; Aksoylar, Aydan
      Patent No.: 9,869,866
      Issue Date: Jan 16, 2018
    • Title: "Method for Determining a Sequence for Drilling Holes According to a Pattern using Global and Local Optimization"
      Inventors: Garaas, Tyler W; Brand, Matthew E.
      Patent No.: 9,703,915
      Issue Date: Jul 11, 2017
    • Title: "MPC Controllers Using Parallel Quadratic Programming"
      Inventors: Di Cairano, Stefano; Brand, Matthew E.
      Patent No.: 9,618,912
      Issue Date: Apr 11, 2017
    • Title: "Method for Generating Representations Polylines Using Piecewise Fitted Geometric Primitives"
      Inventors: Brand, Matthew E.; Marks, Tim; MV, Rohith
      Patent No.: 9,613,443
      Issue Date: Apr 4, 2017
    • Title: "Method for Generating Trajectory for Numerical Control Process"
      Inventors: Brand, Matthew E.; Agrawal, Amit K.; Erdim, Huseyin
      Patent No.: 9,513,623
      Issue Date: Dec 6, 2016
    • Title: "System and Method for Planning a Radiation Therapy Treatment"
      Inventors: Brand, Matthew E.
      Patent No.: 9,251,302
      Issue Date: Feb 2, 2016
    • Title: "Method and System for Cutting Features From Sheet Materials With a Laser Cutter According to a Pattern"
      Inventors: Garaas, Tyler W; Brand, Matthew E.; Josef, Cibulka
      Patent No.: 9,248,525
      Issue Date: Feb 2, 2016
    • Title: "Method for Reconstructing 3D Lines from 2D Lines in an Image"
      Inventors: Ramalingam, Srikumar; Brand, Matthew E.
      Patent No.: 9,183,635
      Issue Date: Nov 10, 2015
    • Title: "Determining Trajectories of Redundant Actuators Jointly Tracking Reference Trajectory"
      Inventors: Shilpiekandula, Vijay; Brand, Matthew E.; Srikanth, Manohar; Bortoff, Scott A.
      Patent No.: 9,170,580
      Issue Date: Oct 27, 2015
    • Title: "System and Method for Controlling Machines According to Pattern of Contours"
      Inventors: Brand, Matthew E.
      Patent No.: 9,104,192
      Issue Date: Aug 11, 2015
    • Title: "Method and System for Detouring Around Features Cut From Sheet Materials with a Laser Cutter According to a Pattern"
      Inventors: Garaas, Tyler W; Brand, Matthew E.
      Patent No.: 9,046,888
      Issue Date: Jun 2, 2015
    • Title: "Method for Scheduling Cars in Elevator Systems to Minimizes Round-Trip Times"
      Inventors: Brand, Matthew E.
      Patent No.: 8,950,555
      Issue Date: Feb 10, 2015
    • Title: "Method for Performing Image Processing Applications Using Quadratic Programming"
      Inventors: Brand, Matthew E.; Chen, Dongui
      Patent No.: 8,761,533
      Issue Date: Jun 24, 2014
    • Title: "Method for Solving Control Problems"
      Inventors: Brand, Matthew E.; Yao, Chen; Shilpiekandula, Vijay
      Patent No.: 8,554,343
      Issue Date: Oct 8, 2013
    • Title: "Method for Optimization Radiotherapy Particle Beams"
      Inventors: Brand, Matthew E.
      Patent No.: 8,492,735
      Issue Date: Jul 23, 2013
    • Title: "Motion Planning for Elevator Cars Moving Independently in One Elevator Shaft"
      Inventors: Brand, Matthew E.
      Patent No.: 8,424,651
      Issue Date: Apr 23, 2013
    • Title: "Motion Planning for Elevator Cars Moving Independently in One Elevator Shaft"
      Inventors: Brand, Matthew E.
      Patent No.: 8,424,650
      Issue Date: Apr 23, 2013
    • Title: "Content Aware Resizing of Images and Videos"
      Inventors: Brand, Matthew E.; Shamir, Ariel; Rubinstein, Michael; Avidan, Shmuel
      Patent No.: 8,380,010
      Issue Date: Feb 19, 2013
    • Title: "Method and System for Localizing in Urban Environments From Omni-Direction Skyline Images"
      Inventors: Ramalingam, Srikumar; Brand, Matthew E.
      Patent No.: 8,311,285
      Issue Date: Nov 13, 2012
    • Title: "Method for Temporally Editing Video"
      Inventors: Brand, Matthew E.
      Patent No.: 8,290,298
      Issue Date: Oct 16, 2012
    • Title: "Method for Editing Images and Videos"
      Inventors: Brand, Matthew E.
      Patent No.: 8,290,297
      Issue Date: Oct 16, 2012
    • Title: "Method for Determining a Location From Images Acquired of an Environment with an Omni-Directional Camera"
      Inventors: Ramalingam, Srikumar; Brand, Matthew E.; Bouaziz, Sofien
      Patent No.: 8,249,302
      Issue Date: Aug 21, 2012
    • Title: "Method and Apparatus for Touching-Up Images"
      Inventors: Brand, Matthew E.; Pletscher, Patrick A.
      Patent No.: 8,160,396
      Issue Date: Apr 17, 2012
    • Title: "Resource Allocation for Rateless Transmissions"
      Inventors: Brand, Matthew E.
      Patent No.: 8,155,048
      Issue Date: Apr 10, 2012
    • Title: "Method for Routing Packets in Wireless Ad-Hoc Networks withProbabilistic Delay Guarantees"
      Inventors: Molisch, Andreas F.; Brand, Matthew E.; Maymounkov, Petar B.
      Patent No.: 8,040,810
      Issue Date: Oct 18, 2011
    • Title: "Method for Routing Packets in Ad-Hoc Networks with Partial Channel State Information"
      Inventors: Molisch, Andreas F.; Brand, Matthew E.
      Patent No.: 7,822,029
      Issue Date: Oct 26, 2010
    • Title: "Method for Finding Minimal Cost Paths under Uncertainty"
      Inventors: Nikolova, Evdokia V.; Brand, Matthew E.
      Patent No.: 7,756,021
      Issue Date: Jul 13, 2010
    • Title: "Method and System for Determining Instantaneous Peak Power Consumption in Elevator Banks"
      Inventors: Brand, Matthew E.; Nikovski, Daniel N.
      Patent No.: 7,743,890
      Issue Date: Jun 29, 2010
    • Title: "Method for Finding Optimal Paths Using a Stochastic NetworkModel"
      Inventors: Mitzenmacher, Michael D.; Brand, Matthew E.; Nikolova, Evdokia V.
      Patent No.: 7,573,866
      Issue Date: Aug 11, 2009
    • Title: "System and Method for Scheduling Elevator Cars Using Pairwise Delay Minimization"
      Inventors: Nikovski, Daniel N.; Brand, Matthew E.; Ebner, Dietmar
      Patent No.: 7,546,905
      Issue Date: Jun 16, 2009
    • Title: "System and Method for Scheduling Elevator Cars Using Branch-and-Bound"
      Inventors: Brand, Matthew E.; Nikovski, Daniel N.; Ebner, Dietmar
      Patent No.: 7,484,597
      Issue Date: Feb 3, 2009
    • Title: "On-Line Recommender System"
      Inventors: Brand, Matthew E.
      Patent No.: 7,475,027
      Issue Date: Jan 6, 2009
    • Title: "Method for Generating a Low-Dimensional Representation of High-Dimensional Data"
      Inventors: Brand, Matthew E.
      Patent No.: 7,412,098
      Issue Date: Aug 12, 2008
    • Title: "Incremental Singular Value Decomposition of Incomplete Data"
      Inventors: Brand, Matthew E.
      Patent No.: 7,359,550
      Issue Date: Apr 15, 2008
    • Title: "Variable Multilinear Models for Facial Synthesis"
      Inventors: Brand, Matthew E.
      Patent No.: 7,133,048
      Issue Date: Nov 7, 2006
    • Title: "Method and System for Scheduling Cars in Elevator Systems Considering Existing and Future Passengers"
      Inventors: Brand, Matthew E.; Nikovski, Daniel N.
      Patent No.: 7,014,015
      Issue Date: Mar 21, 2006
    • Title: "Method for Determining Poses of Sensors"
      Inventors: Brand, Matthew E.
      Patent No.: 7,006,944
      Issue Date: Feb 28, 2006
    • Title: "Modeling Shapes, Motions, Flexions and Textures of Non-Rigid 3D Objects Directly from Video"
      Inventors: Brand, Matthew E.
      Patent No.: 7,006,683
      Issue Date: Feb 28, 2006
    • Title: "Method for Mapping High-Dimensional Samples to Reduced-Dimensional Manifolds"
      Inventors: Brand, Matthew E.
      Patent No.: 6,947,042
      Issue Date: Sep 20, 2005
    • Title: "Rendering Deformable 3D Models Recovered from Videos"
      Inventors: Brand, Matthew E.
      Patent No.: 6,873,724
      Issue Date: Mar 29, 2005
    • Title: "Analysis, Synthesis and Control of Data Signals with Temporal Textures Using a Linear Dynamic System"
      Inventors: Brand, Matthew E.
      Patent No.: 6,864,897
      Issue Date: Mar 8, 2005
    • Title: "Optimal Parking of Free Cars in Elevator Group Control"
      Inventors: Brand, Matthew E.; Nikovski, Daniel N.
      Patent No.: 6,808,049
      Issue Date: Oct 26, 2004
    • Title: "Method for Generating Realistic Facial Animation Directly from Speech Utilizing Hidden Markov Models"
      Inventors: Brand, Matthew E.
      Patent No.: 6,735,566
      Issue Date: May 11, 2004
    • Title: "Method and System for Dynamic Programming of Elevators for Optimal Group Elevator Control"
      Inventors: Brand, Matthew E.; Nikovski, Daniel N.
      Patent No.: 6,672,431
      Issue Date: Jan 6, 2004
    • Title: "Method for Acquiring Static and Dynamic Super-Resolution Texture Maps from Video"
      Inventors: Brand, Matthew E.
      Patent No.: 6,650,335
      Issue Date: Nov 18, 2003
    • Title: "Method for Designing Optimal Single Pointer Predictive Keyboards and Apparatus Therefore"
      Inventors: Brand, Matthew E.
      Patent No.: 6,646,572
      Issue Date: Nov 11, 2003
    • Title: "Method for Predicting Keystroke Characters on Single Pointer Keyboards and Apparatus Therefore"
      Inventors: Brand, Matthew E.
      Patent No.: 6,621,424
      Issue Date: Sep 16, 2003
    • Title: "Method for Inferring Target Paths from Related Cue Paths"
      Inventors: Brand, Matthew E.
      Patent No.: 6,459,808
      Issue Date: Oct 1, 2002
    • Title: "System for Having Concise Models from a Signal Utilizing a Hidden Markov Model"
      Inventors: Brand, Matthew E.
      Patent No.: 6,212,510
      Issue Date: Apr 3, 2001
    • Title: "Markov Model Discriminator Using Negative Examples"
      Inventors: Brand, Matthew E.
      Patent No.: 6,112,021
      Issue Date: Aug 29, 2000
    See All Patents for MERL