Implementing the multilayer retinal model on the complex-cell CNN-UM chip prototype

Dávid Bálya, István Petrás, T. Roska, Ricardo Carmona, Angel Rodríguez Vázquez

Research output: Contribution to journalArticle

21 Citations (Scopus)

Abstract

There is hardly a more important sensory modality for humans and other mammals than vision. The first and best-known part of the visual system is the retina, which is not a mere photoreceptor or static camera but a sophisticated feature preprocessor with continuous input and several parallel output channels. These interacting channels represent the visual scene. Never before has it been known in neuroscience how these channels build up a "visual language". Our mammalian retina model can generate the elements of this visual language. In the present paper the design steps of the implementation of the multilayer CNN retinal model is shown. It is rare that an analogic CNN algorithm has such a sophisticated series of different complex dynamics, mean-while it is feasible on a recently fabricated complex cell CNN-UM chip. The mammalian retina model is decomposed into a full-custom mixed-signal chip that embeds digitally programmable analog parallel processing and distributed image memory on a common silicon substrate. The chip was designed and manufactured in a standard 0.5 μm CMOS technology and contains approximately 500,000 transistors. It consists of 1024 processing units arranged into a 32 × 32 grid. The functional features of the chip are in accordance with the second-order complex cell CNN-UM architecture: two CNN layers with programmable inter- and intra-layer connections between cells as well as programmable layer time constants. The uniqueness of this approach, among others, lies in the reprogrammability, i.e. the openness to any new discovery, even after a possible retinal implementation.

Original languageEnglish
Pages (from-to)427-451
Number of pages25
JournalInternational Journal of Bifurcation and Chaos in Applied Sciences and Engineering
Volume14
Issue number2
DOIs
Publication statusPublished - Feb 2004

Fingerprint

Cell Complex
Visual languages
Retina
Multilayer
Multilayers
Chip
Prototype
Visual Languages
Mammals
Processing
Visual System
Neuroscience
Transistors
Complex Dynamics
Parallel Processing
Time Constant
Cameras
Model
Modality
Data storage equipment

Keywords

  • CACE1k
  • Cellular neural network
  • Model decomposition
  • Multi-channel inner retina
  • Rabbit retina
  • Silicon retina

ASJC Scopus subject areas

  • General
  • Applied Mathematics

Cite this

Implementing the multilayer retinal model on the complex-cell CNN-UM chip prototype. / Bálya, Dávid; Petrás, István; Roska, T.; Carmona, Ricardo; Vázquez, Angel Rodríguez.

In: International Journal of Bifurcation and Chaos in Applied Sciences and Engineering, Vol. 14, No. 2, 02.2004, p. 427-451.

Research output: Contribution to journalArticle

Bálya, Dávid ; Petrás, István ; Roska, T. ; Carmona, Ricardo ; Vázquez, Angel Rodríguez. / Implementing the multilayer retinal model on the complex-cell CNN-UM chip prototype. In: International Journal of Bifurcation and Chaos in Applied Sciences and Engineering. 2004 ; Vol. 14, No. 2. pp. 427-451.
@article{015cb82ec78c4412bfbc0c3143dacb02,
title = "Implementing the multilayer retinal model on the complex-cell CNN-UM chip prototype",
abstract = "There is hardly a more important sensory modality for humans and other mammals than vision. The first and best-known part of the visual system is the retina, which is not a mere photoreceptor or static camera but a sophisticated feature preprocessor with continuous input and several parallel output channels. These interacting channels represent the visual scene. Never before has it been known in neuroscience how these channels build up a {"}visual language{"}. Our mammalian retina model can generate the elements of this visual language. In the present paper the design steps of the implementation of the multilayer CNN retinal model is shown. It is rare that an analogic CNN algorithm has such a sophisticated series of different complex dynamics, mean-while it is feasible on a recently fabricated complex cell CNN-UM chip. The mammalian retina model is decomposed into a full-custom mixed-signal chip that embeds digitally programmable analog parallel processing and distributed image memory on a common silicon substrate. The chip was designed and manufactured in a standard 0.5 μm CMOS technology and contains approximately 500,000 transistors. It consists of 1024 processing units arranged into a 32 × 32 grid. The functional features of the chip are in accordance with the second-order complex cell CNN-UM architecture: two CNN layers with programmable inter- and intra-layer connections between cells as well as programmable layer time constants. The uniqueness of this approach, among others, lies in the reprogrammability, i.e. the openness to any new discovery, even after a possible retinal implementation.",
keywords = "CACE1k, Cellular neural network, Model decomposition, Multi-channel inner retina, Rabbit retina, Silicon retina",
author = "D{\'a}vid B{\'a}lya and Istv{\'a}n Petr{\'a}s and T. Roska and Ricardo Carmona and V{\'a}zquez, {Angel Rodr{\'i}guez}",
year = "2004",
month = "2",
doi = "10.1142/S0218127404009302",
language = "English",
volume = "14",
pages = "427--451",
journal = "International Journal of Bifurcation and Chaos in Applied Sciences and Engineering",
issn = "0218-1274",
publisher = "World Scientific Publishing Co. Pte Ltd",
number = "2",

}

TY - JOUR

T1 - Implementing the multilayer retinal model on the complex-cell CNN-UM chip prototype

AU - Bálya, Dávid

AU - Petrás, István

AU - Roska, T.

AU - Carmona, Ricardo

AU - Vázquez, Angel Rodríguez

PY - 2004/2

Y1 - 2004/2

N2 - There is hardly a more important sensory modality for humans and other mammals than vision. The first and best-known part of the visual system is the retina, which is not a mere photoreceptor or static camera but a sophisticated feature preprocessor with continuous input and several parallel output channels. These interacting channels represent the visual scene. Never before has it been known in neuroscience how these channels build up a "visual language". Our mammalian retina model can generate the elements of this visual language. In the present paper the design steps of the implementation of the multilayer CNN retinal model is shown. It is rare that an analogic CNN algorithm has such a sophisticated series of different complex dynamics, mean-while it is feasible on a recently fabricated complex cell CNN-UM chip. The mammalian retina model is decomposed into a full-custom mixed-signal chip that embeds digitally programmable analog parallel processing and distributed image memory on a common silicon substrate. The chip was designed and manufactured in a standard 0.5 μm CMOS technology and contains approximately 500,000 transistors. It consists of 1024 processing units arranged into a 32 × 32 grid. The functional features of the chip are in accordance with the second-order complex cell CNN-UM architecture: two CNN layers with programmable inter- and intra-layer connections between cells as well as programmable layer time constants. The uniqueness of this approach, among others, lies in the reprogrammability, i.e. the openness to any new discovery, even after a possible retinal implementation.

AB - There is hardly a more important sensory modality for humans and other mammals than vision. The first and best-known part of the visual system is the retina, which is not a mere photoreceptor or static camera but a sophisticated feature preprocessor with continuous input and several parallel output channels. These interacting channels represent the visual scene. Never before has it been known in neuroscience how these channels build up a "visual language". Our mammalian retina model can generate the elements of this visual language. In the present paper the design steps of the implementation of the multilayer CNN retinal model is shown. It is rare that an analogic CNN algorithm has such a sophisticated series of different complex dynamics, mean-while it is feasible on a recently fabricated complex cell CNN-UM chip. The mammalian retina model is decomposed into a full-custom mixed-signal chip that embeds digitally programmable analog parallel processing and distributed image memory on a common silicon substrate. The chip was designed and manufactured in a standard 0.5 μm CMOS technology and contains approximately 500,000 transistors. It consists of 1024 processing units arranged into a 32 × 32 grid. The functional features of the chip are in accordance with the second-order complex cell CNN-UM architecture: two CNN layers with programmable inter- and intra-layer connections between cells as well as programmable layer time constants. The uniqueness of this approach, among others, lies in the reprogrammability, i.e. the openness to any new discovery, even after a possible retinal implementation.

KW - CACE1k

KW - Cellular neural network

KW - Model decomposition

KW - Multi-channel inner retina

KW - Rabbit retina

KW - Silicon retina

UR - http://www.scopus.com/inward/record.url?scp=4344663372&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=4344663372&partnerID=8YFLogxK

U2 - 10.1142/S0218127404009302

DO - 10.1142/S0218127404009302

M3 - Article

AN - SCOPUS:4344663372

VL - 14

SP - 427

EP - 451

JO - International Journal of Bifurcation and Chaos in Applied Sciences and Engineering

JF - International Journal of Bifurcation and Chaos in Applied Sciences and Engineering

SN - 0218-1274

IS - 2

ER -