Cellular Nonlinear Networks: State of the Art and Applications - Systematic Design for Emergence in Cellular Nonlinear Networks - page 13

comes naturally by arranging the cells given the technology constraint. Often the
arrays are two-dimensional with each cell connected to the immediate neighbors.
In digital technologies cellular computers can be realized either as dedicated
VLSI products but in this case they are less flexible and have large development
costs or better using the FPGA (field-programmable-gate-array) technology which
offers a high degree of flexibility and programmability as a results of allowing
random interconnection between a number of identical yet reconfigurable digital
cells. Several FPGA implementations of cellular systems were reported so far
[35], including a CNN prototyping board. Digital technologies have the disadvan-
tage of a low density of cells.
This is why an alternative solution is the use of mixed-signal cells , where com-
pact cells based on nonlinear computation in analog devices are used. The CNN
paradigm was from the beginning oriented to such a technology. So far a series of
chips (today called visual microprocessors or imagers) was developed. Among the
latest implementation solution of a CNN universal machine is the ACE16k chip
[36] which has optical input and can implement the standard CNN model. It has a
number of about 16,000 cells, each cell being associated with an image pixel. Cer-
tain high-speed processing tasks were demonstrated such as the task of classifying
objects (medical pills) presented to the optical input with a speed of 20,000 objects
per second. Several other mixed-signal cell architectures were proposed, for
instance the architecture described in [37], which is based on a nested (recurrent)
utilization of a nonmonotone nonlinear function to reduce the complexity and
implement arbitrary Boolean functions. Arbitrary Boolean functions with n inputs
can be implemented with a linear complexity in n. A schematic of this function is
presented in Chap. 3.
A novel implementation medium, although yet experimental, is represented by
the use of nano-technologies [38]. Browsing the literature of the last 2 years, we
found an increased interest in the “quantum dot” cellular computing paradigm
[39], abbreviated QCA (from quantum cellular automata). Recently [40] the para-
digm evolved into molecular QCA. More details and java simulations of QCA
systems can be found at http://www.nd.edu/~qcahome/. A quantum dot is a nano-
meter scale active cell. Quantum dots are interconnected by proximity so there is
no need for additional metal layers as in standard electronic technologies. A state
change in a QCA cell propagates to its neighbors and this is exactly the basis for
cellular computation. When arranged properly such cells are capable to do various
computational tasks.