Szczegóły publikacji
Opis bibliograficzny
Supervised training of siamese spiking neural networks with Earth Mover's Distance / Mateusz PABIAN, Dominik RZEPKA, Mirosław PAWLAK // W: ICASSP 2022 [Dokument elektroniczny] : 2022 IEEE International Conference on Acoustics, Speech, and Signal Processing : 7–13 May 2022, virtual, 22–27 May 2022, Singapore, satellite venue: Shenzhen, China : proceedings. — Wersja do Windows. — Dane tekstowe. — Piscataway : The Institute of Electrical and Electronics Engineers, cop. 2022. — (Proceedings of the ... IEEE International Conference on Acoustics, Speech, and Signal Processing ; ISSN 1520-6149). — e-ISBN: 978-1-6654-0540-9. — S. 4233–4237. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 4236–4237, Abstr.
Autorzy (3)
Słowa kluczowe
Dane bibliometryczne
ID BaDAP | 144944 |
---|---|
Data dodania do BaDAP | 2023-01-31 |
Tekst źródłowy | URL |
DOI | 10.1109/ICASSP43922.2022.9746630 |
Rok publikacji | 2022 |
Typ publikacji | materiały konferencyjne (aut.) |
Otwarty dostęp | |
Wydawca | Institute of Electrical and Electronics Engineers (IEEE) |
Konferencja | 2022 IEEE International Conference on Acoustics, Speech and Signal Processing |
Czasopismo/seria | Proceedings of the ... IEEE International Conference on Acoustics, Speech, and Signal Processing |
Abstract
This study adapts the highly-versatile siamese neural network model to the event data domain. We introduce a supervised training framework for optimizing Earth Mover's Distance (EMD) between spike trains with spiking neural networks (SNN). We train this model on images of the MNIST dataset converted into spiking domain with novel conversion schemes. The quality of the siamese embeddings of input images was evaluated by measuring the classifier performance for different dataset coding types. The models achieved performance similar to existing SNN-based approaches (F1-score of up to 0.9386) while using only about 15% of hidden layer neurons to classify each example. Furthermore, models which did not employ a sparse neural code were about 45% slower than their sparse counterparts. These properties make the model suitable for low energy consumption and low prediction latency applications.