2001 · 21 citations · 1 references
EngineeringComputer EngineeringElectrical EngineeringArtificial IntelligenceComputer ArchitectureMicroelectronicsCmos Circuit DesignDigital Integrated Circuit DesignDigital Integrated CircuitsDigital Signal ProcessingNeural NetworksActivation FunctionsCircuit ComputationEquivalent Circuit AnalysisEquivalent Circuit ModelingFloating-point ArithmeticFloating-point Unit DesignHigh-order Polynomial ApproximationPiecewise Polynomial Approximation
In this paper we propose a polynomial approximation of the sigmoid activation function and its derivative used in artificial neural networks, and we describe the design of the equivalent digital circuit using a floating-point representation for numbers. The simulation of the circuit realized with CMOS technology AMS 0.35/spl mu/m under a frequency of 300 MHz shows the efficiency of the implementation.
1