Figure
1.1: General model of Stochastic computing system
Design of Stochastic system is generally illustrated in
Fig.1.1. In this model, two parts
which are expressed as two domains are Conventional computing and
Stochastic computing. First, quantities in conventional format
will be scaled into the interval [0, 1] for unipolar
representation. Then Stochastic number generators (SNGs) will be
used to generate stochastic sequences from the scaled inputs. Note that
these sequences carry informations which represent the inputs. At the
moment, our work is to design Stochastic system’s architecture including
the Stochastic operators (SOs) and their connections which will
be used for stochastic algorithms. Finally, the received results from
counters are represented into interval [0, 1], then
traditional encoding will be obtained by using a reverse transformation.