Page 55 - Data Science Algorithms in a Week
P. 55

Using Deep Learning to Configure Parallel Distributed Discrete-Event Simulators  39

                       Input and Output Vectors for a Sample

                          This research implements cognitive weights to measure the complexity of a parallel
                       discrete  event  simulation  with  respect  of  implemented  algorithms.  Because  each
                       simulation object in a simulation implements discrete events defined as code functions,
                       the  complexity  of  each  object  is  also  computed  by  applying  equation  11  for  all
                       event/methods  mapped  to  each  simulation  object.  As  a  result,  several  parameters  that
                       gage simulation complexity are then used as inputs to the deep belief neural network for
                       deep  learning.  These  are:  Total  Simulation  program  cognitive  weights,  maximum
                       cognitive weights of all simulation objects, minimum cognitive weights of all simulation
                       objects, mean cognitive weights of all objects.
                          In  addition,  we  have  captured  other  parameters  that  define  the  hardware,  flow
                       processing, potential messaging and other important characteristics that define a parallel
                       distributed  discrete-event  simulator  implementation.  The  different  components  are
                       defined as follows:

                              1.  Total  Simulation  Program  Cognitive  Weights:  It  is  the  total  number  of
                                 cognitive weights of the simulation program.
                              2.  Number of Simulation objects: It is the total number of simulation objects
                                 in the simulation.
                              3.  Types  of  Simulation  objects:  It  is  the  number  of  classes  of  Simulation
                                 Objects utilized in the simulation.
                              4.  Mean  Events  per  Simulation  Object:  It  is  the  mean  of  the  events  per
                                 simulation object.
                              5.  STD  Events  per  Simulation  Object:  It  is  the  standard  deviation  of  the
                                 events per simulation object.
                              6.  Mean of Cognitive Weights of All objects: It is the mean of the number of
                                 cognitive weights used by the simulation objects in the simulation.
                              7.  STD Cognitive Weights of All objects: It is the standard deviation of the
                                 number of cognitive weights used by the simulation objects in the simulation.
                              8.  Number  of  Global  Nodes:  It  is  the  total  number  of  Global  Nodes  in  the
                                 simulation.
                              9.  Mean  Local  Nodes  per Computer:  It  is  the  mean  of the local  nodes  per
                                 global node utilized in the simulation.
                              10. STD Local Nodes per Computer: It is the standard deviation of the local
                                 nodes per global node utilized in the simulation.
                              11. Mean Number of cores: It is the mean number of cores/threads utilized by
                                 each global node in the simulation.
                              12. STD  Number  of  cores:  It  is  the  standard  deviation  of  number  of
                                 cores/threads utilized by each global node in the simulation.
                              13. Mean processor Speed: It is the mean processor speed of the CPUs used in
                                 the simulation.
   50   51   52   53   54   55   56   57   58   59   60