Title :
Dynamically coupled multi-layer mixed-mode CNN
Author :
Laiho, Mika ; Paasio, Ari
Author_Institution :
Microelectron. Lab., Turku Univ., Finland
Abstract :
This paper shows a multi-layer cellular nonlinear network with feedback interconnections between layers. The multi-layer approach allows for imitation of some complex spatio-temporal visual information processing tasks that are motivated by recent biological studies. A mixed-mode architecture with discrete time integration is used, which makes it possible to process the data layer by layer and still achieve inter-layer feedback interaction. It is shown how the integration step can be programmed differently for each layer with the mixed-mode cell. Also, an in-cell analog-to-digital converter ADC is described and the selected multiplication scheme, the use of binary weighted multipliers, is described.
Keywords :
analogue-digital conversion; cellular neural nets; computer vision; multilayer perceptrons; neural chips; recurrent neural nets; analog-to-digital converter; binary weighted multipliers; discrete time integration; dynamically coupled mixed-mode CNN; feedback interconnections; in-cell analog-to-digital converter ADC; inter-layer feedback interaction; multi-layer cellular nonlinear network; spatio-temporal visual information processing; Analog-digital conversion; Cellular networks; Cellular neural networks; Feedback; Hardware; Information processing; Laboratories; Microelectronics; Nonhomogeneous media; Spatiotemporal phenomena;
Conference_Titel :
Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on
Print_ISBN :
0-7803-8834-8
DOI :
10.1109/ISCAS.2005.1465959