DOCSLIB.ORG

Publications of Sebastian Seung

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Publications of Sebastian Seung Publications of Sebastian Seung V. Jain, H. S. Seung, and S. C. Turaga. Machines that learn to segment images: a crucial technology for connectomics. (pdf) Curr. Opin. Neurobiol (2010). V. Jain, B. Bollmann, M. Richardson, D. Berger et al. Boundary learning by optimization with topological constraints. (pdf supplementary info) Proceedings of the IEEE 23rd Conference on Computer Vision and Pattern Recognition (2010). I. R. Wickersham, H. A. Sullivan, and H. S. Seung. Production of glycoprotein-deleted rabies viruses for monosynaptic tracing and high-level gene expression in neurons. (pdf) Nature Protocols. 5, 596-606 (2010). S. C. Turaga, J. F. Murray, V. Jain, F. Roth, M. Helmstaedter, K. Briggman, W. Denk, and H. S. Seung. Convolutional networks can learn to generate affinity graphs for image segmentation. (pdf) Neural Computation 22, 511-538 (2010). S. C. Turaga, K. L. Briggman, M. Helmstaedter, W. Denk, and H. S. Seung. Maximin affinity learning of image segmentation. (pdf) Adv. Neural Info. Proc. Syst. 22 (Proceedings of NIPS '09) (2010). J. Wang, M. T. Hasan, and H. S. Seung. Laser-evoked synaptic transmission in cultured hippocampal neurons expressing Channelrhodopsin-2 delivered by adeno-associated virus.(pdf) J. Neurosci. Methods 183, 165-175 (2009). Y. Loewenstein, D. Prelec, and H. S. Seung. Operant matching as a Nash equilibrium of an intertemporal game. (pdf) Neural Computation 21, 2755-2773 (2009). H. S. Seung. Reading the Book of Memory: Sparse Sampling versus Dense Mapping of Connectomes. (pdf) Neuron 62, 17-29 (2009). V. Jain and H. S. Seung. Natural Image Denoising with Convolutional Networks (pdf) Adv. Neural Info. Proc. Syst. 21 (Proceedings of NIPS '08) 769-776 (2009). V. Jain, J. F. Murray, F. Roth, S. Turaga, V. Zhigulin, K. L. Briggman, M. N. Helmstaedter, W. Denk, and H. S. Seung. Supervised Learning of Image Restoration with Convolutional Networks.(pdf) Proceedings: IEEE 11th International Conference on Computer Vision (ICCV) (2007) C. Fang-Yen, M. C. Chu, H. S. Seung, R. R. Dasari, and M. S. Feld. Phase-referenced probe interferometer for biological surface profiling and displacement measurements.(pdf) Rev. Sci. Instrum. 78, 123703 (2007). I. R. Fiete, M. S. Fee, and H. S. Seung. Model of birdsong learning based on gradient estimation by dynamic perturbation of neural conductances.(pdf) J. Neurophysiol. 98, 2038-57 (2007). C. Fang-Yen, S. Oh, Y. Park, W. Choi, S. Song, H. S. Seung, R. R. Dasari, and M. S. Feld. Imaging voltage-dependent cell motions with heterodyne Mach-Zehnder phase microscopy. Opt. Lett. 32, 1572-4 (2007). U. Rokni, A. G. Richardson, E. Bizzi, and H. S. Seung. Motor learning with unstable neural representations.(pdf) Neuron 54, 65366 (2007). D. Z. Jin, F. M. Ramazanoglu, and H. S. Seung. Intrinsic bursting enhances the robustness of a neural network model of sequence generation by avian brain area HVC.(pdf) J. Comput. Neurosci. 23, 283-299 (2007; Epub 2007 Apr 18). Y. Loewenstein and H. S. Seung. Operant matching is a generic outcome of synaptic plasticity based on the covariance between reward and neural activity. (pdf) Proc. Natl. Acad. Sci. USA 103, 15224-15229 (2006). I. R. Fiete and H. S. Seung. Gradient learning in spiking neural networks by dynamic perturbation of conductances. (pdf) Phys. Rev. Lett. 97, 048104 (2006). V. Jain, V. Zhigulin, and H. S. Seung. Representing part-whole relationships in recurrent neural networks. (pdf) Adv. Neural Info. Proc. Syst. 18, 563--70 (2006). G. S. Corrado, L. P. Sugrue, H. S. Seung, and W. T. Newsome. Linear-nonlinear-Poisson models of primate choice dynamics. (pdf) J. Exp. Anal. Behav. 84, 581-617 (2005). D. Z. Jin, V. Dragoi, M. Sur, and H. S. Seung. Tilt aftereffect and adaptation-induced changes in orientation tuning in visual cortex. (pdf) J. Neurophysiol. 94, 4038-4050 (2005). R. Tedrake, T. W. Zhang, and H. S. Seung. Learning to Walk in 20 Minutes. (pdf) In Proceedings of the Fourteenth Yale Workshop on Adaptive and Learning Systems, Yale University, New Haven, CT, 2005. J. Werfel, X. Xie, and H. S. Seung. Learning curves for stochastic gradient descent in linear feedforward networks. (pdf) Neural Comput. 17, 2699-2718 (2005). Earlier version: J. Werfel, X. Xie, and H. S. Seung. Learning curves for stochastic gradient descent in linear feedforward networks. (pdf) Adv. Neural Info. Proc. Syst. 16 (2004). A. Starovoytov, J. Choi, and H. S. Seung. Light-directed electrical stimulation of neurons cultured on silicon wafers. (pdf) J. Neurophysiol. 93, 1090-1098 (2005; Epub 2004 Sep 22). C. Fang-Yen, M. C. Chu, H. S. Seung, R. R. Dasari, and M. S. Feld. Noncontact measurement of nerve displacement during action potential with a dual-beam low-coherence interferometer. (pdf) Opt. Lett. 29, 2028-2030 (2004). B. D. Mensh, J. Werfel, and H. S. Seung. BCI Competition 2003 - Data set Ia: combining gamma- band power with slow cortical potentials to improve single-trial classification of electroencephalographic signals. (pdf) IEEE Trans. Biomed. Eng. 51, 1052-1056 (2004). I. R. Fiete, R. H. R. Hahnloser, M. S. Fee, and H. S. Seung. Temporal sparseness of the premotor drive is important for rapid learning in a neural network model of birdsong. (pdf) J. Neurophysiol. 92, 2274-2282 (2004). G. Major, R. Baker, E. Aksay, H. S. Seung, and D. W. Tank. Plasticity and tuning of the time course of analog persistent firing in a neural integrator. (pdf) Proc. Natl. Acad. Sci. USA 101, 7745-7750 (2004). G. Major, R. Baker, E. Aksay, B. Mensh, H. S. Seung, and D. W. Tank. Plasticity and tuning by visual feedback of the stability of a neural integrator. (pdf) Proc. Natl. Acad. Sci. 101, 7739-7744 (2004). X. Xie and H. S. Seung. Learning in neural networks by reinforcement of irregular spiking. (pdf) Phys. Rev. E69, 041909 (2004). B. D. Mensh, E. Aksay, D. D. Lee, H. S. Seung, and D. W. Tank. Spontaneous eye movements in goldfish: Oculomotor integrator performance, plasticity, and dependence on visual feedback. (pdf) Vision Res. 44, 711-726 (2004). R. Tedrake, T. W. Zhang, and H. S. Seung. Stochastic Policy Gradient Reinforcement Learning on a Simple 3D Biped. (pdf) In Proceedings of the IEEE International Conference on Intelligent Robots and Systems (IROS), pages 2849-2854, Sendai, Japan, September 2004. R. Tedrake, T. W. Zhang, M.-F. Fong, and H. S. Seung. Actuating a Simple 3D Passive Dynamic Walker. (pdf) In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), volume 5, pages 4656-4661, New Orleans, LA, April 2004. H. S. Seung. Learning in spiking neural networks by reinforcement of stochastic synaptic transmission. (pdf) Neuron 40, 1063-1073 (2003). E. Aksay, R. Baker, H. S. Seung, and D. W. Tank. Correlated discharge among cell pairs within the oculomotor horizontal velocity-to-position integrator. (pdf) J. Neuroscience 23, 10852-10858 (2003). M. S. Goldman, J. H. Levine, G. Major, D. W. Tank, and H. S. Seung. Robust Persistent Neural Activity in a Model Integrator with Multiple Hysteretic Dendrites per Neuron. (pdf) Cerebral Cortex 13, 1185-1195 (2003). E. Aksay, G. Major, M. S. Goldman, R. Baker, H. S. Seung, and D. W. Tank. History dependence of rate covariation between neurons during persistent activity in an oculomotor integrator. (pdf) Cerebral Cortex 13, 1173-1184 (2003). H. S. Seung. Amplification, Attenuation, and Integration (pdf) in The Handbook of Brain Theory and Neural Networks: Second Edition (M. A. Arbib, Editor) Cambridge, MA: MIT Press, pp. 94-97 (2003). R. H. R. Hahnloser, H. S. Seung, and J. J. Slotine. Permitted and forbidden sets in symmetric threshold-linear networks. (pdf) Neural Comput. 15, 621-38 (2003). Earlier version: R. H. R. Hahnloser, H. S. Seung. Permitted and Forbidden Sets in Symmetric Threshold Linear Networks. (pdf) Adv. Neural Info. Proc. Syst. 13, 217-223 (2001). X. Xie and H. S. Seung. Equivalence of backpropagation and contrastive Hebbian learning in a layered network. (pdf) Neural Comput. 15, 441-54 (2003). X. Xie, R. H. R. Hahnloser, and H. S. Seung. Double-ring network modeling of the head-direction system (pdf) Phys. Rev. E66, 041902 (2002). X. Xie, R. H. R. Hahnloser, and H. S. Seung. Selectively grouping neurons in recurrent networks of lateral inhibition. (pdf) Neural Comput. 14, 2627-46 (2002). Earlier version: X.-H. Xie, R. Hahnloser and H. S. Seung. Learning winner-take-all competition between groups of neurons in lateral inhibitory networks (pdf) Adv. Neural Info. Proc. Syst. 13, 350-356 (2001). M. S. Goldman, C. R. Kaneko, G. Major, E. Aksay, D. W. Tank, and H. S. Seung. Linear regression of eye velocity on eye position and head velocity suggests a common oculomotor neural integrator. (pdf) J. Neurophysiol. 88, 659-65 (2002). D. Z. Jin and H. S. Seung. Fast computation with spikes in a recurrent neural network. (pdf) Phys. Rev. E65, 051922 (2002). R. Tedrake and H. S. Seung. Improved Dynamic Stability using Reinforcement Learning. (pdf) In 5th International Conference on Climbing and Walking Robots and the Support Technologies for Mobile Machines (CLAWAR), pages 341-348, Paris, France, September 2002. E. Aksay, G. Gamkrelidze, H. S. Seung, R. Baker, and D. W. Tank. In vivo intracellular recording and perturbation of persistent activity in a neural integrator (pdf) Nature Neurosci. 4, 184- 93 (2001). D. D. Lee and H. S. Seung. Algorithms for non-negative matrix factorization. (pdf) Adv. Neural Info. Proc. Syst. 13, 556-562 (2001). H. S. Seung and D.D. Lee. The Manifold ways of perception. (pdf) Science 290, 2268-69 (2000). H. S. Seung. Half a century of Hebb. (pdf) Nature Neurosci. 3, 1166-67 (2000). R. H. R. Hahnloser, R. Sarpeshkar, M. A. Mahowald, R. J. Douglas, and H. S. Seung. Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit.
Load more

Recommended publications
  • Spiking Allows Neurons to Estimate Their Causal Effect
    bioRxiv preprint doi: https://doi.org/10.1101/253351; this version posted January 25, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license. Spiking allows neurons to estimate their causal effect Benjamin James Lansdell1,+ and Konrad Paul Kording1 1Department of Bioengineering, University of Pennsylvania, PA, USA [email protected] Learning aims at causing better performance and the typical gradient descent learning is an approximate causal estimator. However real neurons spike, making their gradients undefined. Interestingly, a popular technique in economics, regression discontinuity design, estimates causal effects using such discontinuities. Here we show how the spiking threshold can reveal the influence of a neuron's activity on the performance, indicating a deep link between simple learning rules and economics-style causal inference. Learning is typically conceptualized as changing a neuron's properties to cause better performance or improve the reward R. This is a problem of causality: to learn, a neuron needs to estimate its causal influence on reward, βi. The typical solution linearizes the problem and leads to popular gradient descent- based (GD) approaches of the form βGD = @R . However gradient descent is just one possible approximation i @hi to the estimation of causal influences. Focusing on the underlying causality problem promises new ways of understanding learning. Gradient descent is problematic as a model for biological learning for two reasons. First, real neurons spike, as opposed to units in artificial neural networks (ANNs), and their rates are usually quite slow so that the discreteness of their output matters (e.g.
    [Show full text]
  • Glia and Neurons Glia and Neurons Objective: Role of Glia in Nervous System Agenda: 1
    Applied Neuroscience Columbia Science Honors Program Spring 2017 Glia and Neurons Glia and Neurons Objective: Role of Glia in Nervous System Agenda: 1. Glia • Guest Lecture by Dr. Jennifer Ziegenfuss 2. Machine Learning • Applications in Neuroscience Connectomics A connectome is a comprehensive map of neural connections in the brain (wiring diagram). • Fails to illustrate how neurons behave in real-time (neural dynamics) • Fails to show how a behavior is generated • Fails to account for glia A Tiny Piece of the Connectome Serial EM Reconstruction of Axonal Inputs (various colors) onto a section of apical dendrite (grey) of a pyramidal neuron in mouse cerebral cortex. Arrows mark functional synapses. Lichtman Lab (Harvard) Scaling up efforts to map neural connections is unlikely to deliver the promised benefits such as understanding: • how the brain produces memories • Perception • Consciousness • Treatments for epilepsy, depression, and schizophrenia While glia have been neglected in the quest to understand neuronal signaling, they can sense neuronal activity and control it. Are Glia the Genius Cells? Structure-Function Divide in Brain The function of a neural network is critically dependent upon its interconnections. Only C. elegans has a complete connectome. In neuroscience: • many common diseases and disorders have no known histological trace • debates how many cell types there are • questionable plan to image whole volumes by EM • complexity of structure How is the brain’s function related to its complex structure? Worm Connectome Dots are individual neurons and lines represent axons. C. elegans Caenorhabditis elegans (C. elegans) is a transparent nematode commonly used in neuroscience research. They have a simple nervous system: 302 neurons and 7000 synapses.
    [Show full text]
  • UNDERSTANDING the BRAIN Tbook Collections
    FROM THE NEW YORK TIMES ARCHIVES UNDERSTANDING THE BRAIN TBook Collections Copyright © 2015 The New York Times Company. All rights reserved. Cover Photograph by Zach Wise for The New York Times This ebook was created using Vook. All of the articles in this work originally appeared in The New York Times. eISBN: 9781508000877 The New York Times Company New York, NY www.nytimes.com www.nytimes.com/tbooks Obama Seeking to Boost Study of Human Brain By JOHN MARKOFF FEB. 17, 2013 The Obama administration is planning a decade-long scientific effort to examine the workings of the human brain and build a comprehensive map of its activity, seeking to do for the brain what the Human Genome Project did for genetics. The project, which the administration has been looking to unveil as early as March, will include federal agencies, private foundations and teams of neuroscientists and nanoscientists in a concerted effort to advance the knowledge of the brain’s billions of neurons and gain greater insights into perception, actions and, ultimately, consciousness. Scientists with the highest hopes for the project also see it as a way to develop the technology essential to understanding diseases like Alzheimer’sand Parkinson’s, as well as to find new therapies for a variety of mental illnesses. Moreover, the project holds the potential of paving the way for advances in artificial intelligence. The project, which could ultimately cost billions of dollars, is expected to be part of the president’s budget proposal next month. And, four scientists and representatives of research institutions said they had participated in planning for what is being called the Brain Activity Map project.
    [Show full text]
  • Application of Stochastic Modeling Techniques to Long-Term, High-Resolution Time-Lapse Microscopy Of
    Quantitative analysis of axon elongation: A novel application of stochastic modeling techniques to long-term, high-resolution time-lapse microscopy of cortical axons MASSACHUSETTS INSTTilTE OF TECHNOLOGY by APR 0 1 2010 Neville Espi Sanjana LIBRARIES B.S., Symbolic Systems, Stanford University (2001) A.B., English, Stanford University (2001) Submitted to the Department of Brain & Cognitive Sciences in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Neuroscience at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY February 2010 @ Massachusetts Institute of Technology 2010. All rights reserved. A u th or ....... ...................................... Department of Brain & Cognitive Sciences December 17, 2009 Certified by..... H. Sebastian Seung Professor of Computational Neuroscience - 40" / Thesis Supervisor Accepted by............................ Earl . Miller Earl K. Miller Picower Professor of Neuroscience Chairman, Department Committee on Graduate Theses Quantitative analysis of axon elongation: A novel application of stochastic modeling techniques to long-term, high-resolution time-lapse microscopy of cortical axons by Neville Espi Sanjana Submitted to the Department of Brain & Cognitive Sciences on December 17, 2009, in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Neuroscience Abstract Axons exhibit a rich variety of behaviors, such as elongation, turning, branching, and fas- ciculation, all in service of the complex goal of wiring up the brain. In order to quantify these behaviors, I have developed a system for in vitro imaging of axon growth cones with time-lapse fluorescence microscopy. Image tiles are automatically captured and assembled into a mosaic image of a square millimeter region. GFP-expressing mouse cortical neurons can be imaged once every few minutes for up to weeks if phototoxicity is minimized.
    [Show full text]
  • An Error Detection and Correction Framework for Connectomics
    An Error Detection and Correction Framework for Connectomics Jonathan Zung Ignacio Tartavull∗ Kisuk Leey Princeton University Princeton University MIT [email protected] [email protected] [email protected] H. Sebastian Seung Princeton University [email protected] Abstract We define and study error detection and correction tasks that are useful for 3D reconstruction of neurons from electron microscopic imagery, and for image seg- mentation more generally. Both tasks take as input the raw image and a binary mask representing a candidate object. For the error detection task, the desired output is a map of split and merge errors in the object. For the error correction task, the desired output is the true object. We call this object mask pruning, because the candidate object mask is assumed to be a superset of the true object. We train multiscale 3D convolutional networks to perform both tasks. We find that the error-detecting net can achieve high accuracy. The accuracy of the error-correcting net is enhanced if its input object mask is “advice” (union of erroneous objects) from the error-detecting net. 1 Introduction While neuronal circuits can be reconstructed from volumetric electron microscopic imagery, the process has historically [39] and even recently [37] been highly laborious. One of the most time- consuming reconstruction tasks is the tracing of the brain’s “wires,” or neuronal branches. This task is an example of instance segmentation, and can be automated through computer detection of the boundaries between neurons. Convolutional networks were first applied to neuronal boundary detection a decade ago [14, 38]. Since then convolutional nets have become the standard approach, arXiv:1708.02599v2 [cs.CV] 3 Dec 2017 and the accuracy of boundary detection has become impressively high [40, 3, 21, 9].
    [Show full text]
  • Seymour Knowles-Barley1*, Mike Roberts2, Narayanan Kasthuri1, Dongil Lee1, Hanspeter Pfister2 and Jeff W. Lichtman1
    Mojo 2.0: Connectome Annotation Tool Seymour Knowles-Barley1*, Mike Roberts2, Narayanan Kasthuri1, Dongil Lee1, Hanspeter Pfister2 and Jeff W. Lichtman1 1Harvard University, Department of Molecular and Cellular Biology, USA 2Harvard University, School of Engineering and Applied Sciences, USA Available online at http://www.rhoana.org/ Introduction A connectome is the wiring diagram of connections in a nervous system. Mapping this network of connections is necessary for discovering the underlying architecture of the brain and investigating the physical underpinning of cognition, intelligence, and consciousness [1, 2, 3]. It is also an important step in understanding how connectivity patterns are altered by mental illnesses, learning disorders, and age related changes in the brain. Fully automatic computer vision techniques are available to segment electron microscopy (EM) connectome data. Currently we use the Rhoana pipeline to process images, but results are still far from perfect and require additional human annotation to produce an accurate connectivity map [4]. Here we present Mojo 2.0, an open source, interactive, scalable annotation tool to correct errors in automatic segmentation results. Mojo Features ● Interactive annotation tool ● Smart scribble interface for split and merge operations ● Scalable up to TB scale volumes ● Entire segmentation pipeline including The Mojo interface displaying an EM section from mouse cortex, approximately 7.5x5μm. Mojo is open source and available online: http://www.rhoana.org/ Annotation Results A mouse brain cortex dataset was annotated with the Rhoana image processing pipeline and corrected using Mojo. Partial annotations over two sub volumes totaling 2123μm3 were made by novice users. On average, 1μm3 was annotated in 15 minutes and required 126 edit operations.
    [Show full text]
  • Connectome Seymour Knowles-Barley1,2, Thouis R
    CATEGORY: COMPUTER VISION - CV04 POSTER CONTACT NAME P4212 Thouis Jones: [email protected] Deep Learning for the Connectome Seymour Knowles-Barley1,2, Thouis R. Jones1,2, Josh Morgan2, Dongil Lee2, Narayanan Kasthuri2, Jeff W. Lichtman2 and Hanspeter Pfister1. 1Harvard University, School of Engineering and Applied Sciences, USA. 2Harvard University, Department of Molecular and Cellular Biology, USA. Introduction The connectome is the wiring diagram of the nervous system. Mapping this network is necessary for discovering the underlying architecture of the brain and investigating the physical underpinning of cognition, intelligence, and consciousness [1, 2, 3]. It is also an important step in understanding how connectivity patterns are altered by mental illnesses, learning disorders, and age related changes in the brain. In our experiments, brain tissue is microtomed and imaged with an electron microscope at a resolution of 4x4x30nm per voxel. Cell membranes and sub-cellular structures such as mitochondria, synapses and vesicles are clearly visible in the images. We have developed a fully automatic segmentation pipelines [2] to annotate this data. The requirement of a very high level of accuracy and data rates approaching 1TB a day make this a challenging computer vision task. Deep Learning The first and most critical stage of the our segmentation pipeline (www.rhoana.org) is to identify membrane in the 2D images produced by the microscope. We use convolutional neural networks to classify images, trained on a few hand-annotated images. GPU computing greatly reduces the training and classification time for these computationally demanding networks. In this work, we used pylearn2 [5] to apply maxout networks [6] to two connectome data sets.
    [Show full text]
  • Cryonics Magazine, September-October, 2012
    September-October 2012 • Volume 33:5 Register for Alcor’s 40th Anniversary Conference Page 17 Symposium on Cryonics and Brain- Threatening Disorders Page 8 Book Review: Connectome: How the Brain’s Wiring Makes Us Who We Are Page 12 ISSN 1054-4305 $9.95 Improve Your Odds of a Good Cryopreservation You have your cryonics funding and contracts in place but have you considered other steps you can take to prevent problems down the road? _ Keep Alcor up-to-date about personal and medical changes. _ Update your Alcor paperwork to reflect your current wishes. _ Execute a cryonics-friendly Living Will and Durable Power of Attorney for Health Care. _ Wear your bracelet and talk to your friends and family about your desire to be cryopreserved. _ Ask your relatives to sign Affidavits stating that they will not interfere with your cryopreservation. _ Attend local cryonics meetings or start a local group yourself. _ Contribute to Alcor’s operations and research. Contact Alcor (1-877-462-5267) and let us know how we can assist you. Take a look at the Your source for news about: ALCOR BLOG • Cryonics technology • Cryopreservation cases • Television programs about cryonics http://www.alcor.org/blog/ • Speaking events and meetings • Employment opportunities Alcor Life Connect with Alcor members and supporters on our Extension official Facebook page: http://www.facebook.com/alcor.life.extension. Foundation foundation is on Become a fan and encourage interested friends, family members, and colleagues to support us too. CONTENTS 5 Quod incepimus lume 33:5
    [Show full text]
  • Modularity and Neural Coding from a Brainstem Synaptic Wiring Diagram
    bioRxiv preprint doi: https://doi.org/10.1101/2020.10.28.359620; this version posted October 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-ND 4.0 International license. 1 Modularity and neural coding from a brainstem synaptic 2 wiring diagram 1,† 2, 1,3 4 1,5 1 3 Ashwin Vishwanathan , Alexandro D. Ramirez ,* Jingpeng Wu *, Alex Sood , Runzhe Yang , Nico Kemnitz , Dodam Ih1, Nicholas Turner1,5, Kisuk Lee1,6, Ignacio Tartavull1, William M. Silversmith1, Chris S. Jordan1 Celia David1, Doug Bland1, Mark S. Goldman4, Emre R. F. Aksay2, H. Sebastian Seung1,5,†, the EyeWirers. 4 1. Princeton Neuroscience Institute, Princeton University, Princeton, NJ, USA. ; 2. Institute for Computational Biomedicine and the 5 Department of Physiology and Biophysics, Weill Cornell Medical College, New York, New York, USA, NY.; 3. Center for Computational 6 Neuroscience, Flatiron Institute, NY, USA.; 4. Center for Neuroscience, University of California at Davis, Davis, California, USA.; 5. 7 Computer Science Department, Princeton University, Princeton, NJ, USA.; 6. Brain & Cognitive Sciences Department, Massachusetts 8 Institute of Technology, Cambridge, MA, USA. 9 † Correspondence to [email protected] and [email protected] 10 Abstract 11 Neuronal wiring diagrams reconstructed from electron microscopic images are enabling new ways of attack- 12 ing neuroscience questions. We address two central issues, modularity and neural coding, by reconstructing 13 and analyzing a wiring diagram from a larval zebrafish brainstem. We identified a recurrently connected “cen- 14 ter” within the 3000-node graph, and applied graph clustering algorithms to divide the center into two modules 15 with stronger connectivity within than between modules.
    [Show full text]
  • At the Centre of a New Reality
    Artificial intelligence index At the centre of a new reality The leader of Samsung’s artificial-intelligence research outlines the quest for convenient and natural interaction with machines. By Leigh Dayton ebastian Seung’s new role at Samsung’s is to make users’ interactions with their devices Medicine Catalyst in 2019 (go.nature.com/ R&D Campus in the South Korean cap- “as convenient and natural as possible”. That 2jpnkjn), the app, a collaboration between ital of Seoul is a world away from his may mean turning on the air-conditioning unit Samsung and health-care provider and insurer, previous post at Princeton University with a voice command, running health checks Kaiser Permanente, significantly improves the in New Jersey. Appointed as the head through a smartwatch, or directing robot vac- likelihood that patients will complete rehabil- Sof Samsung Research in June 2020, Seung is uum cleaners with simplified sign language. itation programmes. “The completion rate leading thousands of people in 15 research “Multimodal interaction is core to what we jumped to almost 90%, up from 50, and the centres, who are investigating technologies do,” says Seung. “It has become a buzzword rate of rehospitalization plummeted after such as computer vision, augmented reality, driving AI research at Samsung.” Multimodal HeartWise,” says Seung. “This is a way to save robotics and 5G communications networks. interaction refers to devices that offer the user lives and also help to reduce health-care costs.” Seung also heads seven global artificial intel- multiple modes of interaction, such as sight, ligence (AI) research centres in South Korea, speech and touch.
    [Show full text]
  • Signature Redacted Thes
    Towards Optical Connectomics: Feasibility of 3D Reconstruction of Neural Morphology Using Expansion Microscopy and In Situ Molecular Barcoding Peilun Dai B.Eng. Electrical and Electronic Engineering Nanyang Technological University, 2014 Submitted to the Department of Brain and Cognitive Sciences in Partial Fulfillment of the Requirements for the Degree of Master of Science in Neuroscience at the Massachusetts Institute of Technology September 2018 C 2018 Massachusetts Institute of Technology. All rights reserved. Signature redacted Signature of author: Department ofBrairad-eg1ive ciences Signature redacted August 10, 2018 Certified by: Edward S. Boyden, III Y. Eva Tan Professor in Neurotechnology Associate Professor of Media Arts and Sciences, -Braifne Sciences, and Biological Engineering Thes Accepted by Signature redacted Matthew A. Wilson Sherman Fairchild Professor of Neroscience and Picower Scholar Director of Graduate Education for Brain and Cognitive Sciences MASSACHUSEFS INSTITE OF TECHNOLOGY OCT 112018 LIBRARIES ARCHIVES 2 Towards Optical Connectomics: Feasibility of 3D Reconstruction of Neural Morphology Using Expansion Microscopy and In Situ Molecular Barcoding by Peilun Dai Submitted to the Department of Brain and Cognitive Sciences on August 10, 2018 in Partial Fulfillment of the Requirements for the Degree of Master of Science in Neuroscience Abstract Reconstruction of the 3D morphology of neurons is an essential step towards the analysis and understanding of the structures and functions of neural circuits. Optical microscopy has been a key technology for mapping brain circuits throughout the history of neuroscience due to its low cost, wide usage in biological sciences and ability to read out information-rich molecular signals. However, conventional optical microscopes have limited spatial resolution due to the diffraction limit of light, requiring a tradeoff between the density of neurons to be imaged and the ability to resolve finer structures.
    [Show full text]
  • Efficient Automatic 3D-Reconstruction of Branching Neurons from EM Data
    Efficient Automatic 3D-Reconstruction of Branching Neurons from EM Data Jan Funke1 Bjoern Andres2 Fred A. Hamprecht2 Albert Cardona1 Matthew Cook1 1Institute of Neuroinformatics 2HCI/IWR University Zurich/ETH Zurich University of Heidelberg http://www.ini.uzh.ch http://hci.iwr.uni-heidelberg.de Abstract We present an approach for the automatic reconstruction of neurons from 3D stacks of electron microscopy sections. The core of our system is a set of possible assignments, each of which proposes with some cost a link between neuron re- gions in consecutive sections. These can model the con- tinuation, branching, and end of neurons. The costs are trainable on positive assignment samples. An optimal and consistent set of assignments is found for the whole volume at once by solving an integer linear program. This set of assignments determines both the segmentation into neuron regions and the correspondence between such regions in (a) (b) (c) (d) neighboring slices. For each picked assignment, a confi- dence value helps to prioritize decisions to be reviewed by Figure 1. Overview of our approach. For all slices in the volume a human expert. (a), different segmentation hypotheses are created (b). All possible assignments between these hypotheses are enumerated (c) and, ac- We evaluate the performance of our method on an an- cording to assignment specific costs (obtained by a trainable clas- notated volume of neural tissue and compare to the current sifier), a global optimum of both segmentations and assignments state of the art [26]. Our method is superior in accuracy is found (d). and can be trained using a small number of samples.
    [Show full text]