MULTILAYER NEURON NETWORK AND NETWORK DESIGNING METHOD THEREOF

PURPOSE: To attain high-speed operating by providing a multilayer neural circuit network, with which a link weight value is defined as a constant, having a stage function. CONSTITUTION: This network has an input layer L0 for receiving neurons 31-N33 and the input of (m) bits, hidden layers L1 and L2...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	KAN KOUCHIN, TEI KOUSEN
Format:	Patent
Sprache:	eng
Schlagworte:	ANALOGUE COMPUTERS BASIC ELECTRIC ELEMENTS CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING COUNTING ELECTRIC DIGITAL DATA PROCESSING ELECTRIC SOLID STATE DEVICES NOT OTHERWISE PROVIDED FOR ELECTRICITY INFORMATION STORAGE PHYSICS SEMICONDUCTOR DEVICES STATIC STORES
Online-Zugang:	Volltext bestellen
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	PURPOSE: To attain high-speed operating by providing a multilayer neural circuit network, with which a link weight value is defined as a constant, having a stage function. CONSTITUTION: This network has an input layer L0 for receiving neurons 31-N33 and the input of (m) bits, hidden layers L1 and L2 more than one, matrix-shaped input synapse layer, transmission synapse groups more than one, and bias synapse group. The link weight value of the input synapse group is initialized, and a desired output value is presented for the input and output of input and output layers. The real output is generated, the change component of the weight value is stored, a new weight value is found in respect to the current weight value, and the weight value is decreased. Thus, the multi-layer neural circuit network accelerated more than the conventional network is obtained.