An Artificial Neural Network (ANN) involves a complex network of interconnected nodes called artificial neurons (AN); the AN sums N weighted inputs and send thought the result to a non-linear activation function (AF). In this work, a modified version of the sigmoid activation function is proposed. To obtain a voltage-to-voltage (V - V) transfer function required by an specific ANN. The proposed solution uses a pseudo-differential pair configuration at the input as voltage to current converter. The proposed circuit is designed using a commercial PDK in 180nm (TSMC) and is simulated in Virtuoso (Cadence). This specific design enable to obtain the desired steepness of the sigmoid function by means of the proper transistor sizing. Simulation results of our specific design show that we can reach an average relative error of only 1.09 % for steepness of 1 as compared to the exact mathematical function, and a power consumption of 6.77μW for steepness of 10.

Voltage-to-Voltage Sigmoid Neuron Activation Function Design for Artificial Neural Networks

Crupi F.;Lanuzza M.;
2022-01-01

Abstract

An Artificial Neural Network (ANN) involves a complex network of interconnected nodes called artificial neurons (AN); the AN sums N weighted inputs and send thought the result to a non-linear activation function (AF). In this work, a modified version of the sigmoid activation function is proposed. To obtain a voltage-to-voltage (V - V) transfer function required by an specific ANN. The proposed solution uses a pseudo-differential pair configuration at the input as voltage to current converter. The proposed circuit is designed using a commercial PDK in 180nm (TSMC) and is simulated in Virtuoso (Cadence). This specific design enable to obtain the desired steepness of the sigmoid function by means of the proper transistor sizing. Simulation results of our specific design show that we can reach an average relative error of only 1.09 % for steepness of 1 as compared to the exact mathematical function, and a power consumption of 6.77μW for steepness of 10.
978-1-6654-2008-2
Artificial Neural Network
non-linear transfer function
pseudo differential pair
sigmoid activation function
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11770/338059
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact