|
@@ -34,8 +34,8 @@ Comparison of three heuristics for generating points set}
|
|
|
One of the most important discrepancy notions is the so-called star
|
|
|
discrepancy. Roughly speaking, a point set of low star discrepancy value
|
|
|
allows for a small approximation error in quasi-Monte Carlo integration.
|
|
|
- In this work we present a tool realizing the implantation of three
|
|
|
- basics heuristics for construction low discrepancy points sets
|
|
|
+ In this work we present a tool realizing the implementation of three
|
|
|
+ basics heuristics for the construction of low discrepancy points sets
|
|
|
in the generalized Halton model: fully random search, local search with
|
|
|
simulated annealing and genetic $(5+5)$ search with a ad-hoc
|
|
|
crossover function.
|
|
@@ -57,7 +57,7 @@ presented in figure~\ref{class_flow}.
|
|
|
Experiments were conducted on two machines:
|
|
|
\begin{itemize}
|
|
|
\item 2.4 GHz Intel Dual Core i5 hyper-threaded to 2.8GHz, 8 Go 1600 MHz DDR3.
|
|
|
- \item 2.7 GHz Intel Quad Core i74800MQ hyper-threaded to 3.7GHz, 16 Go 1600 MHz DDR3.
|
|
|
+ \item 2.7 GHz Intel Quad Core i7 4800MQ hyper-threaded to 3.7GHz, 16 Go 1600 MHz DDR3.
|
|
|
\end{itemize}
|
|
|
|
|
|
\begin{figure}
|
|
@@ -72,14 +72,14 @@ Experiments were conducted on two machines:
|
|
|
\label{class_flow}
|
|
|
\end{figure}
|
|
|
|
|
|
-On these machines, some basic profiling has make clear that
|
|
|
+On these machines, some basic profiling has made clear that
|
|
|
the main bottleneck of the computations is hiding in the \emph{computation
|
|
|
of the discrepancy}. The chosen algorithm and implantation of this
|
|
|
cost function is the DEM-algorithm~\cite{Dobkin} of
|
|
|
\emph{Magnus Wahlstr\o m}~\cite{Magnus}.\medskip
|
|
|
|
|
|
All the experiments has been conducted on dimension 2,3, and 4
|
|
|
---- with a fixed Halton basis 7, 13, 29, and 3 ---. Some minor tests have
|
|
|
+--- with a fixed Halton prime basis 7, 13, 29, and 3. Some minor tests have
|
|
|
been made in order to discuss the dependency of the discrepancy and
|
|
|
efficiency of the heuristics with regards to the values chosen for the
|
|
|
prime base. The average results remains roughly identical when taking
|
|
@@ -96,10 +96,17 @@ times in order to smooth up the results and obtain more exploitable datas.
|
|
|
Then an arithmetic mean of the result is performed on the values. In addition
|
|
|
extremal values are also given in order to construct error bands graphs.
|
|
|
|
|
|
+Graph are presented not with the usual box plots to show the
|
|
|
+error bounds, but in a more graphical way with error bands. The graph
|
|
|
+of the mean result is included inside a band of the same color which
|
|
|
+represents the incertitude with regards to the values obtained.
|
|
|
+
|
|
|
A flowchart of the conduct of one experiment is described in the
|
|
|
flowchart~\ref{insight_flow}. The number of iteration of the heuristic is
|
|
|
-I and the number of full restart is N. The function Heuristic() correspond to
|
|
|
-a single step of the chosen heuristic. We now present an in-depth view of
|
|
|
+I and the number of full restart is N. The function Heuristic() corresponds to
|
|
|
+a single step of the chosen heuristic.
|
|
|
+
|
|
|
+We now present an in-depth view of
|
|
|
the implemented heuristics.
|
|
|
|
|
|
\begin{figure}
|
|
@@ -110,21 +117,16 @@ the implemented heuristics.
|
|
|
\end{mdframed}
|
|
|
\end{figure}
|
|
|
|
|
|
-Graph are presented not with the usual box plot to show the
|
|
|
-error bounds, but in a more graphical way with error bands. The graph
|
|
|
-of the mean result is included inside a band of the same color which
|
|
|
-represents the incertitude with regards to the values obtained.
|
|
|
-
|
|
|
\section{Heuristics developed}
|
|
|
|
|
|
\subsection{Fully random search (Test case)}
|
|
|
The first heuristic implemented is the random search. We generate
|
|
|
random permutations, compute the corresponding sets of Halton points
|
|
|
- and select the best set with regard to its discrepancy.
|
|
|
+ and select the best set with regards to its discrepancy.
|
|
|
The process is wrapped up in the
|
|
|
flowchart~\ref{random_flow}. In order to generate at each step random
|
|
|
permutations, we transform them directly from the previous ones.
|
|
|
- More precisely the permutation is a singleton object which have method
|
|
|
+ More precisely the permutation is a singleton object which have a method
|
|
|
random, built on the Knuth Fisher Yates shuffle. This algorithm allows
|
|
|
us to generate an uniformly chosen permutation at each step. We recall
|
|
|
this fact and detail the algorithm in the following section.
|
|
@@ -170,7 +172,7 @@ the naive implementation.
|
|
|
an edge $(i,j)$ exists in the graph if and only if the swap of $T[i]$ and
|
|
|
$T[j]$ had been executed. This graph encodes the permutation represented by
|
|
|
$T$. To be able to encode any permutation the considered graph must be
|
|
|
- connected --- in order to allow any pairs of points to be swapped ---.
|
|
|
+ connected --- in order to allow any pairs of points to be swapped.
|
|
|
Since by construction every points is reached by an edge, and that there
|
|
|
exists exactly $n-1$ edges, we can conclude directly that any permutation can
|
|
|
be reached by the algorithm. Since the probability of getting a fixed graph
|
|
@@ -183,20 +185,20 @@ the naive implementation.
|
|
|
\subsubsection{Results and stability}
|
|
|
We first want to analyze the dependence of the results on the number of
|
|
|
iterations of the heuristic, in order to discuss its stability.
|
|
|
-The results are compiled in the figures~\ref{rand_iter2},~\ref{rand_iter3},
|
|
|
+The results are compiled in the figures~\ref{rand_iter2} and~\ref{rand_iter3},
|
|
|
restricted to a number of points between 80 and 180.
|
|
|
-We emphasize on the fact the lots of datas appears on the graphs,
|
|
|
-and the error bands representation make them a bit messy. These graphs
|
|
|
+We emphasize on the fact lot of datas appears on graphs,
|
|
|
+and error bands representation make them a bit messy. These graphs
|
|
|
were made for extensive internal experiments and parameters researches.
|
|
|
-The final wrap up graphs are much more lighter and only presents the best
|
|
|
+The final wrap up graphs are much more lighter and only present the best
|
|
|
results obtained.
|
|
|
-As expected from a fully random search, the error bands are very large for
|
|
|
-low number of iterations ($15\%$ of the value for 400 iterations) and tends
|
|
|
+As expected from a fully random search, error bands are very large for
|
|
|
+low number of iterations ($15\%$ of the value for 400 iterations) and tend
|
|
|
to shrink with a bigger number of iterations (around $5\%$ for 1500 iterations).
|
|
|
This shrinkage is a direct consequence of well known concentrations bounds
|
|
|
(Chernoff and Asuma-Hoeffding).
|
|
|
The average results are quite stable, they decrease progressively with
|
|
|
-the growing number of iterations, but seem to get to a limits after 1000
|
|
|
+the growing number of iterations, but seem to get to a limit after 1000
|
|
|
iterations. This value acts as a threshold for the interesting number of iterations.
|
|
|
As such interesting results can be conducted with \emph{only} 1000 iterations,
|
|
|
without altering too much the quality of the set with regards to its
|