Szczegóły publikacji
Opis bibliograficzny
Retrieving impressions from semantic memory modeled with associative pulsing neural networks / Patryk ORZECHOWSKI, Adrian HORZYK, Janusz A. Starzyk, Jason H. Moore // W: IEEE SSCI 2018 [Dokument elektroniczny] : proceedings of the 2018 IEEE Symposium Series on Computational Intelligence : 18–21 November 2018, Bengaluru, India / ed. Suresh Sundaram. — Wersja do Windows. — Dane tekstowe. — [Piscataway] : IEEE, cop. 2018. — 1 dysk Flash. — Dod. ISBN 978-1-5386-9276-9. — e-ISBN: 978-1-5386-9275-2. — S. 2060–2067. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 2067, Abstr. — P. Orzechowski - pierwsza afiliacja: University of Pennsylvania, USA
Autorzy (4)
- AGHOrzechowski Patryk
- AGHHorzyk Adrian
- Starzyk Janusz A.
- Moore Jason H.
Słowa kluczowe
Dane bibliometryczne
ID BaDAP | 118556 |
---|---|
Data dodania do BaDAP | 2018-12-07 |
DOI | 10.1109/SSCI.2018.8628780 |
Rok publikacji | 2018 |
Typ publikacji | materiały konferencyjne (aut.) |
Otwarty dostęp | |
Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
Konferencja | 2018 IEEE Symposium Series on Computational Intelligence |
Abstract
Associative Pulsing Neuron (APN) is a recently proposed model of neurons which incorporates multiple functional aspects of biologically-inspired Spiking Neuron (SN) while maintaining the simplicity of the simulation model. The APN model also enriches the functionality of SN models with conditional plasticity rules which let APNs to intelligently connect and represent associative relations that can be retrieved from raw training data or deduced from activities of presynaptic neurons. Associative Pulsing Neural Networks (APNNs) built with APN neurons use these plasticity rules to self-organize their network structures, are computationally efficient and provide easy interpretable and intuitive results. In this paper, we model the human-like activity of impression retrieval from a semantic memory using a selforganizing APNN structure spanned on a graph built from a bag of words (BOW), a popular data structure typically used for information retrieval.We illustrate that generation of associations in the network and present suitability of the proposed model to retrieve impressions.