Improving Performance in Neural Networks by Dendrite-Activated Connection
Paper i proceeding, 2025
We introduce a novel computational unit for neural networks featuring multiple biases, challenging the conventional perceptron structure. Designed to emphasize preserving uncorrupted information as it transfers from one unit to the next, this unit applies activation functions later in the process, incorporating specialized biases for each unit. We posit this unit as an improved design for neural networks and support this with (1) empirical evidence across diverse datasets; (2) a class of functions where this unit utilizes parameters more efficiently; and (3) biological analogies suggesting closer mimicry to natural neural processing. Source code is available at https://github.com/CuriosAI/dac-dev.
Författare
Carlo Metta
Istituto di Scienza e Tecnologie dell'Informazione A. Faedo
Marco Fantozzi
Universita degli Studi di Parma
Andrea Papini
Chalmers, Matematiska vetenskaper, Tillämpad matematik och statistik
Göteborgs universitet
Gianluca Amato
Universita degli Studi Gabriele d'annunzio di Chieti-Pescara
Matteo Bergamaschi
Università di Padova
Andrea Fois
Universita degli Studi di Parma
Silvia Giulia Galfré
Universita di Pisa
Alessandro Marchetti
Universita degli Studi Gabriele d'annunzio di Chieti-Pescara
Michelangelo Vegliò
Universita degli Studi Gabriele d'annunzio di Chieti-Pescara
Maurizio Parton
Universita degli Studi Gabriele d'annunzio di Chieti-Pescara
Francesco Morandin
Universita degli Studi di Parma
Studies in Classification Data Analysis and Knowledge Organization
14318814 (ISSN) 21983321 (eISSN)
133-1419783031847011 (ISBN)
14th Scientific Meeting of the Classification and Data Analysis Group of the Italian Statistical Society, CLADAG 2023
Salerno, Italy,
Salerno, Italy,
Ämneskategorier (SSIF 2025)
Formella metoder
Datorteknik
Beräkningsmatematik
DOI
10.1007/978-3-031-84702-8_15