Log-Nets: Logarithmic Feature-Product Layers Yield More Compact Networks

Philipp Grüning*, Thomas Martinetz, Erhardt Barth

*Corresponding author for this work


We introduce Logarithm-Networks (Log-Nets), a novel bio-inspired type of network architecture based on logarithms of feature maps followed by convolutions. Log-Nets are capable of surpassing the performance of traditional convolutional neural networks (CNNs) while using fewer parameters. Performance is evaluated on the Cifar-10 and ImageNet benchmarks.

Original languageEnglish
Title of host publicationICANN 2020: Artificial Neural Networks and Machine Learning – ICANN 2020
EditorsIgor Farkaš, Paolo Masulli, Stefan Wermter
Number of pages13
Volume12397 LNCS
Place of PublicationCham
PublisherSpringer, Cham
Publication date14.10.2020
ISBN (Print)978-3-030-61615-1
ISBN (Electronic)978-3-030-61616-8
Publication statusPublished - 14.10.2020
Event29th International Conference on Artificial Neural Networks - Bratislava, Slovakia
Duration: 15.09.202018.09.2020
Conference number: 250349

Research Areas and Centers

  • Centers: Center for Artificial Intelligence Luebeck (ZKIL)
  • Research Area: Intelligent Systems

DFG Research Classification Scheme

  • 409-05 Interactive and Intelligent Systems, Image and Language Processing, Computer Graphics and Visualisation


Dive into the research topics of 'Log-Nets: Logarithmic Feature-Product Layers Yield More Compact Networks'. Together they form a unique fingerprint.

Cite this