📄 randomize_extend.html
字号:
<html> <head><title>Custom constraints</title></head><body bgcolor="#ffffff"><h3>Custom constraints</h3>This page describes how to extend the set of modules that implementconstraints, cooling schemes, and permutation schemes for annealed surrogates.I have tried to keep everything as modular as possible with FORTRAN. Therefore,it seems extremely unlikely that, when you supply a new module for any one ofthe three tasks, you might want to change any of the others, or even the mainmodule, <font color=blue><tt>randomize.f</tt></font>. Experts in simulatedannealing are likely to know better cooling and permutation schemes than thoseavailable in this package. Experts, however, are likely to figure out how toprovide these and I will therefore concentrate on the implementation of othercost functions.<p>If you want to supply any changed or new modules, here is the recommended sequence of actions. <ul><li>Choose a new name for the module and place the source in the appropriate directory (<font color=blue><tt>cost</tt></font>, <font color=blue><tt>cool</tt>, or </font><font color=blue><tt>perm</tt></font>)<li>Add a line for the target <font color=blue><tt>all:</tt></font> to the <font color=blue><tt>Makefile</tt></font> in the directory <font color=blue><tt>randomize</tt></font><li>Create the new executable by running <font color=blue><tt>make</tt></font></ul>For example, you could start from <fontcolor=blue><tt>cost/auto.f</tt></font> by adding or deleting constraints andthen name the source <font color=blue><tt>cost/mycost.f</tt></font> Then addthe line <blockquote><font color=blue><tt>$(TEMPLATE) COST=mycost COOL=exp PERM=rand$(XFLAGS)</tt></font> </blockquote>to <font color=blue><tt>Makefile</tt></font> and run<font color=blue><tt>make</tt></font>.<p><hr><p><h4>Cost function entry points</h4>In order for the modularity to work, the subroutine calls must be keptidentical through all constraint/cost function modules. This means that anyadditional information has to be exchanged and stored by means of commonblocks. The blank common block<br><font color=blue><tt>common nmax,cost,temp,cmin,rate,x</tt></font><br>contains some useful items which are likely to be needed by themodules. These are the number <tt>nmax</tt> of data points, the current cost<tt>cost</tt>, the current temperature <tt>temp</tt>, the minimal cost<tt>cmin</tt>, the current rate of acceptance <tt>rate</tt> and the data setas a real array <tt>x</tt>. Note that <tt>x</tt> needs to be of fixed lengthsyntactically, but since it is the last item of the common block, this is notessential as long as it is declared large enough to store <tt>nmax</tt> items.<p>The following entry points must be provided:<p><font color=blue><tt>subroutine opts_cost(ncol)</tt></font><blockquote>Scans the command line for options specific to the cost function. Set <fontcolor=blue><tt>ncol</tt></font> to the number of columns needed from the inputfile. On input, it is set to what is deduced from the<font color=blue><tt> -c </tt></font> and <font color=blue><tt> -m </tt></font>options.<p><font color=red><b>Example: </b></font> in <font color=blue><tt>cost/auto.f</tt></font> the number of desired lags is determined from the command line option <font color=blue><tt> -D </tt></font> and stored as an item in a common block to be available for the actual cost function. Similarly the option <font color=blue><tt> -W </tt></font> is scanned for the desired weighting.</blockquote><p><font color=blue><tt>subroutine what_cost()</tt></font><blockquote>Just prints a line saying which cost function is implemented.</blockquote><p><font color=blue><tt>subroutine usage_cost()</tt></font><blockquote>Prints that part of the usage message specific to the cost function, inparticular maybe explaining the command line options.</blockquote><p><font color=blue><tt>subroutine cost_init()</tt></font><blockquote>Does all necessary initializations of the cost function. In particular, thegoal value of some quantities, typically derived from the original time series,can be set. <p><font color=red><b>Example: </b></font> in <font color=blue><tt>cost/auto.f</tt></font> the number of lags is truncated to fit into the available space. Then the autocorrelation function of the data (<tt>x</tt> in the blank common block) is computed and stored as an array <tt>c0</tt> in the common block <tt>/costcom/</tt> for later use.</blockquote><p><p><font color=blue><tt>subroutine cost_transform(nmax,mcmax,nxdum,x)</tt></font><blockquote>Does any initial transformation on the data. The data length <fontcolor=blue><tt>nmax</tt></font>, the number of columns <fontcolor=blue><tt>mcmax</tt></font>, and the first dimension <fontcolor=blue><tt>nxdum</tt></font> is provided in case that is needed. Finally,<font color=blue><tt>x</tt></font> is the data.<p><font color=red><b>Example: </b></font> in <font color=blue><tt>cost/auto.f</tt></font> the data is normalized to zero mean and unit variance, mean and variance are stored.</blockquote><p><font color=blue><tt>subroutine cost_inverse(nmax,mcmax,nxdum,x,y)</tt></font><blockquote>Inverts any initial transformation on the data. The data length <fontcolor=blue><tt>nmax</tt></font>, the number of columns <fontcolor=blue><tt>mcmax</tt></font>, and the first dimension <fontcolor=blue><tt>nxdum</tt></font> is provided in case that is needed. Finally,<font color=blue><tt>x</tt></font> is the data and <fontcolor=blue><tt>y</tt></font> a back transformed copy.<p><font color=red><b>Example: </b></font> in <font color=blue><tt>cost/auto.f</tt></font> the data is restored to its original mean and variance.</blockquote><p><font color=blue><tt>subroutine cost_full(iv)</tt></font><blockquote>Determines the value of the cost function given a configuration <tt>x</tt>.A nonzero value <tt>iv</tt> indicates that diagnostic output is desired.<p><font color=red><b>Example: </b></font> in <font color=blue><tt>cost/auto.f</tt></font> the autocorrelation function of the data is computed and averaged to yield the total cost, which is returned by the function.</blockquote><p><font color=blue><tt>subroutine cost_update(n1,n2,cmax,iaccept,iv)</tt></font><blockquote>Determines the value of the cost function when the data items <tt>n1</tt> and<tt>n2</tt> are exchanged. If the resulting cost is less than <tt>cmax</tt>,the change is accepted. In that case, <tt>iaccept</tt> is set to one. The newcost is stored in the blank common blok. On nonacceptance, its value is leftunchanged. A nonzero value <tt>iv</tt> indicates that diagnostic output isdesired. <p><font color=red><b>Example: </b></font> it is important to note that while the complete cost function requires O(<tt>nlag*nmax</tt>) computations, the interchange of two items can be monitored by O(<tt>nlag</tt>) computations. This is taken care of in <font color=blue><tt>cost/auto.f</tt></font>. Be absolutely sure that the update is done correctly for all possible cases! A mismatch between assumed cost and actual cost can lead to a desaster. Because of possible roundoff problems, the full computation is called for regularly by the cooling scheme. </blockquote><p> <a href="randomize.html"><em>constrained randomization</em></a> *<a href="../contents.html">Table of Contents</a> * <a href="../../index.html" target="_top">TISEAN home</a></body></html>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -