Activation functions and their impact on the training and performance of convolutional neural network models

Gerald Onwujekwe, Victoria Yoon

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

    Abstract

    Activation functions are a very crucial part of convolutional neural networks (CNN) because to a very large extent, they determine the performance of the CNN model. Various activation functions have been developed over the years and the choice of activation function to use in a given model is usually a matter of trial and error. In this paper, we evaluate some of the most-used activation functions and how they impact the time to train a CNN model and the performance of the model. We make recommendations for the best activation functions to use based on the results of our experiment.

    Original languageEnglish
    Title of host publication26th Americas Conference on Information Systems, AMCIS 2020
    PublisherAssociation for Information Systems
    ISBN (Electronic)9781733632546
    StatePublished - 2020
    Event26th Americas Conference on Information Systems, AMCIS 2020 - Salt Lake City, Virtual, United States
    Duration: Aug 10 2020Aug 14 2020

    Publication series

    Name26th Americas Conference on Information Systems, AMCIS 2020

    Conference

    Conference26th Americas Conference on Information Systems, AMCIS 2020
    Country/TerritoryUnited States
    CitySalt Lake City, Virtual
    Period08/10/2008/14/20

    Keywords

    • Accuracy
    • Activation function
    • Convolutional neural network
    • Performance
    • Training time

    Fingerprint

    Dive into the research topics of 'Activation functions and their impact on the training and performance of convolutional neural network models'. Together they form a unique fingerprint.

    Cite this