On the Circuit Complexity of Feedforward Neural Networks

Valeriu Beiu; Joos A. Peperstraete; Jan Vandewalle; Rudy Lauwereins

doi:10.1007/978-1-4471-2097-1_121

On the Circuit Complexity of Feedforward Neural Networks

Valeriu Beiu, Joos A. Peperstraete, Jan Vandewalle, Rudy Lauwereins

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

In this paper a network will be considered an acyclic graph. It has several input nodes (inputs) and some (at least one) output nodes (outputs). The nodes of the network are characterized by fan-in (the number of incoming edges) and fan-out (the number of outgoing edges), while the network has a certain size (the number of nodes) and depth (the number of edges on the longest input to output path). If with each edge a synaptic weight is associated and each node computes the weighted sum of its inputs to which a nonlinear activation function is then applied (artificial neuron), the network is a neural network (NN). Similarly, a Boolean circuit (BC) is a network of gates implementing elementary Boolean functions (AND, OR, NOT).

Original language	English
Title of host publication	International Conference on Artificial Neural Networks
DOIs	https://doi.org/10.1007/978-1-4471-2097-1_121
Publication status	Published - May 26 1994
Event	ICANN'94 - Sorrento, Italy Duration: May 26 1994 → …

Conference

Conference	ICANN'94
Period	5/26/94 → …

Access to Document

10.1007/978-1-4471-2097-1_121

http://link.springer.com/chapter/10.1007%2F978-1-4471-2097-1_121

Cite this

@inproceedings{2cb6fe5c69b044ccb4b884be8507ab48,

title = "On the Circuit Complexity of Feedforward Neural Networks",

abstract = "In this paper a network will be considered an acyclic graph. It has several input nodes (inputs) and some (at least one) output nodes (outputs). The nodes of the network are characterized by fan-in (the number of incoming edges) and fan-out (the number of outgoing edges), while the network has a certain size (the number of nodes) and depth (the number of edges on the longest input to output path). If with each edge a synaptic weight is associated and each node computes the weighted sum of its inputs to which a nonlinear activation function is then applied (artificial neuron), the network is a neural network (NN). Similarly, a Boolean circuit (BC) is a network of gates implementing elementary Boolean functions (AND, OR, NOT).",

author = "Valeriu Beiu and Peperstraete, \{Joos A.\} and Jan Vandewalle and Rudy Lauwereins",

year = "1994",

month = may,

day = "26",

doi = "10.1007/978-1-4471-2097-1\_121",

language = "English",

isbn = "978-3-540-19887-1",

booktitle = "International Conference on Artificial Neural Networks",

note = "ICANN'94 ; Conference date: 26-05-1994",

}

TY - GEN

T1 - On the Circuit Complexity of Feedforward Neural Networks

AU - Beiu, Valeriu

AU - Peperstraete, Joos A.

AU - Vandewalle, Jan

AU - Lauwereins, Rudy

PY - 1994/5/26

Y1 - 1994/5/26

N2 - In this paper a network will be considered an acyclic graph. It has several input nodes (inputs) and some (at least one) output nodes (outputs). The nodes of the network are characterized by fan-in (the number of incoming edges) and fan-out (the number of outgoing edges), while the network has a certain size (the number of nodes) and depth (the number of edges on the longest input to output path). If with each edge a synaptic weight is associated and each node computes the weighted sum of its inputs to which a nonlinear activation function is then applied (artificial neuron), the network is a neural network (NN). Similarly, a Boolean circuit (BC) is a network of gates implementing elementary Boolean functions (AND, OR, NOT).

AB - In this paper a network will be considered an acyclic graph. It has several input nodes (inputs) and some (at least one) output nodes (outputs). The nodes of the network are characterized by fan-in (the number of incoming edges) and fan-out (the number of outgoing edges), while the network has a certain size (the number of nodes) and depth (the number of edges on the longest input to output path). If with each edge a synaptic weight is associated and each node computes the weighted sum of its inputs to which a nonlinear activation function is then applied (artificial neuron), the network is a neural network (NN). Similarly, a Boolean circuit (BC) is a network of gates implementing elementary Boolean functions (AND, OR, NOT).

U2 - 10.1007/978-1-4471-2097-1_121

DO - 10.1007/978-1-4471-2097-1_121

M3 - Conference contribution

SN - 978-3-540-19887-1

BT - International Conference on Artificial Neural Networks

T2 - ICANN'94

Y2 - 26 May 1994

ER -

On the Circuit Complexity of Feedforward Neural Networks

Abstract

Conference

Access to Document

Fingerprint

Cite this