DOCSLIB.ORG

Based, Distributed Synapses

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Based, Distributed Synapses Research Collection Doctoral Thesis Robust online learning in neuromorphic systems with spike- based, distributed synapses Author(s): Stefanini, Fabio Publication Date: 2013 Permanent Link: https://doi.org/10.3929/ethz-a-010118610 Rights / License: In Copyright - Non-Commercial Use Permitted This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use. ETH Library DISS. ETH NO. 21337 Robust online learning in neuromorphic systems with spike-based, distributed synapses A dissertation submitted to ETH ZURICH for the degree of Doctor of Sciences presented by Fabio Stefanini Institute of Neuroinformatics Laurea di Dottore in Fisica Università degli Studi di Roma “La Sapienza” born April, 19th, 1983 citizen of Italy accepted on the recommendation of Prof. Dr. Rodney Douglas, examiner Prof. Dr. Giacomo Indiveri, co-examiner Prof. Dr. Stefano Fusi, co-examiner 2013 Classification in VLSI ii To my wife. Classification in VLSI iv Abstract Neuromorphic Very Large Scale Integration (VLSI) hardware offers a low-power and compact electronic substrate for implementing distributed plastic synapses for artificial systems. However, the technology used for constructing analog neuromorphic circuits has limited resolution and high intrinsic variability, leading to large circuit mismatch. Consequently, neuromorphic synapse circuits are imprecise and thus the mapping of pre-defined models onto neural processing systems built with such components is practically unfeasible. This problem of variability can be avoided by off-loading learning to ad-hoc digital resources external to the synapse but this work-around introduces communication bottlenecks and so compromises compactness and power efficiency. Here I propose a more direct solution using a system composed of aggregated classifiers and distributed, stochastic learning to exploit the intrinsic variability and mismatch of the substrate. The system consists of a feed-forward neural network with one hidden layer of randomly connected neurons with fixed weights and a pool of binary classifiers with stochastic learning. To demonstrate the system, I developed software procedures to configure the hardware classifier and used them to test its ability to learn to identify a target class. The method follows earlier works where the neuron circuit parameters are first configured to obtain the desired behavior and then the plastic synapses are characterized by imposing determined pre- and post- synaptic mean activities. The classifier is then included in a simulated environment to compare the complete system to theoretical and computational models. In particular, a well-known dataset consisting of images of hand-written digits has been used as benchmark to demonstrate with experimental results that by aggregating the responses of imprecise classifiers trained with the feed-forward network it is possible to achieve a performance which is comparable to state-of-the-art. The intrinsic noise present in the neuromorphic hardware and the type of distributed plastic synapses used result in an effective combination of classifiers, as described by recent machine learning techniques. This work is a crucial step towards enabling online learning through synaptic plasticity on imprecise hardware, thus providing a novel, resource efficient substrate for machine learning. It demonstrates how a “bottom-up” approach that exploits the properties of the substrate to obtain a functioning neuromorphic system can elegantly overcome the complications due to the calibration of imprecise circuits with limited configurability, thus extending the valuability and feasibility of analog VLSI techniques. The biological plausibility of the system links to relevant issues in neuroscience, such as the role of intrinsic variability in probabilistic computation in the brain, and suggests possible design strategies for future emerging technologies. v Classification in VLSI Compendio Hardware neuromorfo costruito con tecnologia VLSI offre un substrato compatto di tipo low-power per l’implementazione di sinapsi plastiche distribuite in sistemi artificiali. Il tipo di tecnologia correntemente utilizzata per realizzare circuiti neuromorfi presenta tuttavia una bassa risoluzione e un’alta variabilità intrinseca, quindi una grande disomogeneità circuitale, cioè mismatch. Per questo motivo i circuiti neuromorfi che realizzano le sinapsi sono imprecisi e una mappatura di modelli prestabiliti su sistemi costruiti a partire da tali componenti diventa una soluzione praticamente impercorribile. Il problema della variabilità può essere evitato dedicando risorse specializzate ai meccanismi di apprendimento, separate dal blocco di circuiti neuro-sinaptici, tuttavia questa opzione introduce colli di bottiglia di comunicazione e compromette compattezza circuitale e efficienza energetica. La mia proposta consiste in una soluzione più diretta che usa un sistema composto da classificatori aggregati e apprendimento stocastico distribuito al fine di sfruttare la variabilità intrinseca e il mismatch del substrato. Il sistema consiste in una rete neurale di tipo feed-forward con un livello intermedio di neuroni connessi a random, con pesi sinaptici fissi, e un insieme di classificatori binari con apprendimento stocastico. Per dimostrare la funzionalità del sistema, ho sviluppato le procedure software necessarie per la configurazione dei classificatori hardware e le ho usati per testare le loro capacità di apprendimento nell’identificare una data classe di stimoli. Questo metodo segue precedenti lavori in cui prima i parameteri circuitali del neurone vengono configurati per ottenere il comportamento desiderato e successivamente una caratterizzazione delle sinapsi plastiche viene ottenuta imponendo determinate attività pre- e post- sinaptiche. Il singolo classificatore è stato poi inserito in una simulazione per confrontare il sistema di classificazione completo con i risultati stimati dalla teoria e dai modelli. In particolare, un set di dati di caratteri scritti a mano comunemente usato in computer science è stato usato come punto di riferimento per dimostrare, con risultati sperimentali, che è possibile ottenere una performance di classificazione confrontabile con lo stato dell’arte tramite una combinazione di risposte di classificatori imprecisi addestrati con la rete feed-forward. La presenza del rumore intrinseco dell’hardware neuromorfo e il tipo di sinapsi impiegate comportano un’efficace combinazione di classificatori, come si trova in certe recenti descrizioni di tecniche di machine learning. Questo lavoro rappresenta un passo in avanti rispetto alla possibilità di realizzare sistemi di apprendimento online sfruttando plasticità sinaptica su hardware con basso fattore di precisione e quindi propone un substrato nuovo, efficiente in termini di utilizzo di risorse, per machine learning. Il lavoro dimostra come un approccio “dal basso” che sfrutti le proprietà del substrato per ottenere un sistema neuromorfo funzionante può superare in maniera elegante le complicazioni dovute alle accurate calibrazioni di hardware impreciso e con limitate possibilità di configurazione e perciò estente la validità e la praticabilità del VLSI analogico. La plausibilità biologica del sistema impiegato si collega a interessanti approfondimenti nelle neuroscienze e suggerisce possibili strategie di design per le future tecnologie emergenti. vi Disclaimer I hereby declare that the work in this thesis is that of the candidate alone, except where indicated in the text and as described below. Chapter 3 is a modified version of the paper [Chicca et al., 2013]. Chapter 4 is a modified version of the paper [Sheik et al., 2011]. The results in appendix summarize the work carried-out in collaboration with Michael Beyeler and Rahel Von Rohr during their Master and Semester projects in the Neuroscience program of University of Zurich and ETH. The network model in Chapter 5 and the classification results in Chapter 6 have been developed in collaboration with Mattia Rigotti (New York University and Columbia University, NYC, New York). The use of “we” in the thesis refers to the aforementioned people in the relevant sections. Publications arising from this thesis The work of this thesis or part of it has been published on journals and conference proceedings as listed below. These publications are also mentioned in the text where relevant. • Sheik, S. and Stefanini, F. and Neftci, E. and Chicca, E. and Indiveri, G., Systematic configuration and automatic tuning of neuromorphic systems, Circuits and Systems (ISCAS), 2011 IEEE International Symposium on, 873-876 (2011) • Indiveri, G. and Stefanini, F. and Chicca, E., Spike-based learning with a generalized integrate and fire silicon neuron, Circuits and Systems (ISCAS), Proceedings of 2010 IEEE International Symposium on, 1951-1954 (2010) • Beyeler, M. and Stefanini, F. and Proske, H. and Galizia, G. and Chicca, E., Exploring olfactory sensory networks: simulations and hardware emulation, Biomedical Circuits and Systems Conference (BioCAS), 2010 IEEE, 270-273 (2010) • E. Neftci, S. Sheik, F. Stefanini, G. Indiveri, A Python package for accessing, configuring, and applying electronic spiking neural networks, Frontiers in Neuroinformatics, Python In Neuroscience, Research Topic, June, 2013, Accepted • S. Sheik, M. Pfeiffer, F. Stefanini, G. Indiveri, Spatio-temporal spike pattern classification in neuromorphic systems, International Conference on Biomimetic and Biohybrid Systems, August, 2013, Accepted • E. Chicca,
Load more

Recommended publications
  • Image Recognition by Spiking Neural Networks
    TALLINN UNIVERSITY OF TECHNOLOGY Faculty of Information Technology Nikolai Jefimov 143689IASM IMAGE RECOGNITION BY SPIKING NEURAL NETWORKS Master's thesis Supervisor: Eduard Petlenkov PhD Tallinn 2017 TALLINNA TEHNIKAÜLIKOOL Infotehnoloogia teaduskond Nikolai Jefimov 143689IASM KUJUTISE TUVASTAMINE IMPULSI NÄRVIVÕRKUDEGA Magistritöö Juhendaja: Eduard Petlenkov PhD Tallinn 2017 Author’s declaration of originality I hereby certify that I am the sole author of this thesis. All the used materials, references to the literature and the work of others have been referred to. This thesis has not been presented for examination anywhere else. Author: Nikolai Jefimov 30.05.2017 3 Abstract The aim of this thesis is to present opportunities on implementations of spiking neural network for image recognition and proving learning algorithms in MATLAB environment. In this thesis provided examples of the use of common types of artificial neural networks for image recognition and classification. Also presented spiking neural network areas of use, as well as the principles of image recognition by this type of neural network. The result of this work is the realization of the learning algorithm in MATLAB environment, as well as the ability to recognize handwritten numbers from the MNIST database. Besides this presented the opportunity to recognizing Latin letters with dimensions of 3x3 and 5x7 pixels. In addition, were proposed User Guide for masters course ISS0023 Intelligent Control System in TUT. This thesis is written in English and is 62 pages long, including 6 chapters, 61 figures and 4 tables. 4 Annotatsioon Kujutise tuvastamine impulsi närvivõrkudega Selle väitekiri eesmärgiks on tutvustada impulsi närvivõrgu kasutamisvõimalused, samuti näidata õpetamise algoritmi MATLAB keskkonnas.
    [Show full text]
  • Table of Contents Welcome from the General Chair
    Table of Contents Welcome from the General Chair . .v Organizing Committee . vi Program Committee . vii-ix IJCNN 2011 Reviewers . ix-xiii Conference Topics . xiv-xv INNS Officers and Board of Governors . xvi INNS President’s Welcome . xvii IEEE CIS Officers and ADCOM . xviii IEEE CIS President’s Welcome . xix Cooperating Societies and Sponsors . xx Conference Information . xxi Hotel Maps . xxii. Schedule-At-A-Glance . xxiii-xxvii Schedule Grids . xxviii-xxxiii IJCNN 2012 Call for Papers . xxxiv-xxxvi Program . 1-31 Detailed Program . 33-146 Author Index . 147 i ii The 2011 International Joint Conference on Neural Networks Final Program July 31 – August 5, 2011 Doubletree Hotel San Jose, California, USA Sponsored by: International Neural Network Society iii The 2011 International Joint Conference on Neural Networks IJCNN 2011 Conference Proceedings © 2011 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redis- tribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE. For obtaining permission, write to IEEE Copyrights Manager, IEEE Operations Center, 445 Hoes Lane, PO Box 1331, Piscataway, NJ 08855-1331 USA. All rights reserved. Papers are printed as received from authors. All opinions expressed in the Proceedings are those of the authors and are not binding on The Institute of Electrical and Electronics Engineers, Inc. Additional
    [Show full text]
  • Neuromorphic Engineering - Wikipedia
    Neuromorphic engineering - Wikipedia https://en.wikipedia.org/wiki/Neuromorphic_engineering Neuromorphic engineering Neuromorphic engineering, also known as neuromorphic computing,[1][2][3] is a concept developed by Carver Mead,[4] in the late 1980s, describing the use of very-large-scale integration (VLSI) systems containing electronic analog circuits to mimic neuro-biological architectures present in the nervous system.[5] In recent times, the term neuromorphic has been used to describe analog, digital, mixed-mode analog/digital VLSI, and software systems that implement models of neural systems (for perception, motor control, or multisensory integration). The implementation of neuromorphic computing on the hardware level can be realized by oxide-based memristors,[6] spintronic memories,[7] threshold switches, and transistors.[8] A key aspect of neuromorphic engineering is understanding how the morphology of individual neurons, circuits, applications, and overall architectures creates desirable computations, affects how information is represented, influences robustness to damage, incorporates learning and development, adapts to local change (plasticity), and facilitates evolutionary change. Neuromorphic engineering is an interdisciplinary subject that takes inspiration from biology, physics, mathematics, computer science, and electronic engineering to design artificial neural systems, such as vision systems, head-eye systems, auditory processors, and autonomous robots, whose physical architecture and design principles are based on those
    [Show full text]
  • The Posthuman Management Matrix: Understanding the Organizational Impact of Radical Biotechnological Convergence
    Part Three Abstract. In this text we present the Posthuman Management Matrix, a model for understanding the ways in which organizations of the future will be af- fected by the blurring – or even dissolution – of boundaries between human beings and computers. In this model, an organization’s employees and con- sumers can include two different kinds of agents (human and artificial) who may possess either of two sets of characteristics (anthropic or computer-like); the model thus defines four types of possible entities. For millennia, the only type of relevance for management theory and practice was that of human agents who possess anthropic characteristics – i.e., natural human beings. During the 20th Century, the arrival of computers and industrial robots made relevant a second type: that of artificial agents possessing computer-like char- acteristics. Management theory and practice have traditionally overlooked the remaining two types of possible entities – human agents possessing computer-like phys- ical and cognitive characteristics (which can be referred to as ‘cyborgs’) and artificial agents possessing anthropic physical and cognitive characteristics (which for lack of a more appropriate term might be called ‘bioroids’) – be- cause such agents did not yet exist to serve as employees or consumers for organizations. However, in this text we argue that ongoing developments in neuroprosthetics, genetic engineering, virtual reality, robotics, and artificial intelligence are indeed giving rise to such types of agents and that new spheres of management theory and practice will be needed to allow organiza- tions to understand the operational, legal, and ethical issues that arise as their pools of potential workers and customers evolve to include human beings whose bodies and minds incorporate ever more computerized elements and artificial entities that increasingly resemble biological beings.
    [Show full text]
  • Engineering Optimization Using Artificial Neural Network
    International Journal of Innovations in Engineering and Technology (IJIET) Engineering Optimization using Artificial Neural Network Ravi Katukam Convener-Engineering Innovation, Cyient Limited, Hyderabad Prajna Behera Graduate Student IIT Kanpur Abstarct - Neural network is one of the important components in Artificial Intelligence (AI). It is a computational model for storing and retrieving acquired knowledge .It has been studied for many years in the hope of achieving human-like performance in many fields, such as speech and image recognition as well as information retrieval. ANNs consists of dense interconnected computing units that are simple models for complex neurons in biological systems. The knowledge is acquired during a learning process and is stored in the synaptic weights of the inter-nodal connections. The main advantage of neural network is their ability to represent complex input/output relationships. The performance of an ANN is measured according to an error between target and actual output, complexity of ANN, training time etc. The topologicalstructure of an ANN can be characterized by the arrangement of the layers and neurons, the nodal connectivity, and the nodal transfer function. This paper contains development of a simple graphical user interface in MATLAB that uses neural network algorithm for prediction of output for a set of inputs according to the learning example given. The developed tool is used to predict material removal rate and tool ware by taking speed, feed and depth of cut as input parameter from a CNC milling machine. This tool is also used for predicting volume and stress of an inboard fixed trailing edge rib for different combination of thickness parameters.
    [Show full text]
  • Coupling Resistive Switching Devices with Neurons: State of the Art and Perspectives
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by Frontiers - Publisher Connector PERSPECTIVE published: 15 February 2017 doi: 10.3389/fnins.2017.00070 Coupling Resistive Switching Devices with Neurons: State of the Art and Perspectives Alessandro Chiolerio 1*, Michela Chiappalone 2, Paolo Ariano 1 and Sergio Bocchini 1 1 Center for Sustainable Future Technologies, Istituto Italiano di Tecnologia, Torino, Italy, 2 Neuroscience and Brain Technologies Department, Istituto Italiano di Tecnologia, Genova, Italy Here we provide the state-of-the-art of bioelectronic interfacing between biological neuronal systems and artificial components, focusing the attention on the potentiality offered by intrinsically neuromorphic synthetic devices based on Resistive Switching (RS). Neuromorphic engineering is outside the scopes of this Perspective. Instead, our focus is on those materials and devices featuring genuine physical effects that could be sought as non-linearity, plasticity, excitation, and extinction which could be directly and more naturally coupled with living biological systems. In view of important applications, such as prosthetics and future life augmentation, a cybernetic parallelism is traced, between biological and artificial systems. We will discuss how such intrinsic features could reduce Edited by: the complexity of conditioning networks for a more natural direct connection between Ramona Samba, biological and synthetic worlds. Putting together living systems with RS devices could Natural and Medical Sciences Institute, Germany represent a feasible though innovative perspective for the future of bionics. Reviewed by: Keywords: cybernetics, bio-electronic systems, resistive switching devices, memristors, neuromorphic devices, Alexander Dityatev, multielectrode arrays German Center for Neurodegenerative Diseases (DZNE), Germany Stefano Vassanelli, INTRODUCTION University of Padua, Italy *Correspondence: The brain is the most powerful and complex known computational system.
    [Show full text]
  • Data Analysis Foundations – Neural Networks Quadrant 1 – E-Text
    e-PGPathshala Subject : Computer Science Paper: Data Analytics Module No 17: CS/DA/17 - Data Analysis Foundations – Neural Networks Quadrant 1 – e-text 1.1 Introduction Neural networks process information in a similar way the human brain does. Neural networks take a different approach to problem solving than that of conventional computers. Artificial neural networks (ANNs) or connectionist systems are a computational model used in machine learning, computer science and other research disciplines, which is based on a large collection of connected simple units called artificial neurons, loosely analogous to axons in a biological brain. Connections between neurons carry an activation signal of varying strength. If the combined incoming signals are strong enough, the neuron becomes activated and the signal travels to other neurons connected to it. Such systems can be trained from examples, rather than explicitly programmed, and excel in areas where the solution or feature detection is difficult to express in a traditional computer program. Like other machine learning methods, neural networks have been used to solve a wide variety of tasks, like computer vision and speech recognition, that are difficult to solve using ordinary rule-based programming. This chapter gives an overview on neural networks. 1.2 Learning Outcomes • Learn the basics of Neural Network • Know applications of Neural Network 1.3 Neural network – The Big Picture Neural network is a computer modeling approach to computation that is loosely based upon the architecture of
    [Show full text]
  • Martina Corazzol Master Thesis
    Aalborg University, Biomedical Engineering and Informatics, Spring Semester, 2012 Master Thesis "Classification of movements of the rat based on intra-cortical signals using artificial neural network and support vector machine" Martina Corazzol Department of Health Science & Technology Frederik Bajers Vej 7D2 9220 Aalborg Telephone (+45) 9940 9940 Fax (+45) 9815 4008 http://www.hst.aau.dk Title: Synopsis: "Classification of movements of the rat based on intra-cortical signals using artifi- cial neural network and support vector ma- Background chine" The goal of intra-cortical brain computer in- terface (BCI) is to restore the lost function- alities in disabled patients suffering from Project period: severely impaired movements. BCIs have the Master Thesis, Spring 2012 advantage to create a direct communication pathway between the brain and an external Project group: device for restoring disability. This is pos- Group 12gr1085 sible by decoding signals from the primary motor cortex and translating them into com- Members: mands for a prosthetic device. The project Martina Corazzol aim was to develop a decoding method based on a rat model. Previously recorded data and an already develop pre-processing method were used. The experimental design Supervisor: was developed starting from intra-cortical Winnie Jensen (IC) signal recorded in the rat primary mo- tor cortex (M1). The data pre-processing included denoising with wavelet technique, Assistant Supervisor: spike detection, and feature extraction. After Sofyan Hammad Himaidan Hammad the firing rates of intra-cortical neurons were extracted, artificial neural network (ANN) and support vector machine (SVM) were ap- plied to classify the rat movements into two No. printed Copies: 4 possible classes, Hit or No Hit.
    [Show full text]