Özet
The traditional neural network topology is not flexible to change during the training process. Every neuron and it's independent weights in the network are part of the solution function. The proposed focusing neuron model utilizes inter-dependent weights produced by a focusing function. This neuron can change it's focus position and aperture. This property allows a flexible-dynamic network topology, which can be trained using conventional back-propagation algorithm. Our experiments show that focusing neuron neural networks achieve higher success than fully connected neural networks.
Tercüme edilen katkı başlığı | Focusing neuron |
---|---|
Orijinal dil | Türkçe |
Ana bilgisayar yayını başlığı | 2017 25th Signal Processing and Communications Applications Conference, SIU 2017 |
Yayınlayan | Institute of Electrical and Electronics Engineers Inc. |
ISBN (Elektronik) | 9781509064946 |
DOI'lar | |
Yayın durumu | Yayınlandı - 27 Haz 2017 |
Harici olarak yayınlandı | Evet |
Etkinlik | 25th Signal Processing and Communications Applications Conference, SIU 2017 - Antalya, Turkey Süre: 15 May 2017 → 18 May 2017 |
Yayın serisi
Adı | 2017 25th Signal Processing and Communications Applications Conference, SIU 2017 |
---|
???event.eventtypes.event.conference???
???event.eventtypes.event.conference??? | 25th Signal Processing and Communications Applications Conference, SIU 2017 |
---|---|
Ülke/Bölge | Turkey |
Şehir | Antalya |
Periyot | 15/05/17 → 18/05/17 |
Bibliyografik not
Publisher Copyright:© 2017 IEEE.
Keywords
- artificial neural network
- focusing neuron