Analog VLSI current mode implementation of artificial neural networks

摘要

In this paper we examine all current-mode building blocks for implementation of artificial neural networks. The building blocks proposed include a current dependent pulse generator, a pulse frequency modulator used to multiply incoming pulses with variable weights, and a current in/current out sigmoid function circuit. The circuits can be easily implemented in a standard CMOS VLSI process. Simulations are presented to confirm the validity of the proposed circuits.