Analog neural network VLSI implementations /
| Main Author: | |
|---|---|
| Other Authors: | , , |
| Format: | Thesis Book |
| Language: | English |
| Published: |
1991.
|
| Subjects: | |
| Online Access: | Link to OAKTrust copy |
| Abstract: | The objective of this dissertation is to demonstrate the viability of using analog circuit design techniques to build neural network systems in hardware. For this we introduce a novel design approach called transconductance-mode (T-mode). It uses transconductance amplifiers and multipliers for gain stages and capacitors to perform integration operations. Using these elements, together with some extra nonlinear resistors, many sets of nonlinear differential equations can be implemented in hardware. The hardware implementation of artificial neural networks can be formulated as a problem of realizing a specific set of nonlinear differential equations. We will show that the proposed T-mode circuit design technique can be used to emulate the differential equations that describe most of the known neural network systems. We will use this technique to build a variety of programmable neural network systems and to implement a learning neural network associative memory with on chip analog dynamic memory. |
|---|---|
| Item Description: | Typescript (photocopy). Vita. "Major subject: Electrical Engineering." |
| Physical Description: | xxv, 325 leaves : illustrations ; 29 cm |
| Bibliography: | Includes bibliographical references. |