COSC-4P82-Final-Project/lib/lilgp/htmlMan/lil-gp.ch7.htm

394 lines
14 KiB
HTML

<HTML>
<HEAD>
<TITLE></TITLE>
</HEAD>
<BODY BGCOLOR="#FFFFFF">
<P><B><FONT SIZE="5">Chapter 7</FONT></B></P>
<PRE> 7.1 <A HREF="#7.1">Tree Representation</A>
7.2 <A HREF="#7.2">Selection Methods</A>
7.3 <A HREF="#7.3">Operators</A>
7.4 <A HREF="#7.4">Miscellany</A>
7.4.1 <A HREF="#7.4.1">Tree Generation Spaces</A>
7.4.2 <A HREF="#7.4.2">Saved Individuals</A>
7.4.3 <A HREF="#7.4.3">Ephemeral Random Constants</A></PRE>
<P>
<HR>
</P>
<P>
<B><FONT SIZE="5">Extending the Kernel<BR>
</FONT></B>
</P>
<P>
Internally, lil-gp is fairly simple. I have attempted to keep
the structure reasonably clean and modular, without going too
overboard about avoiding global variables and such. 24 C files
comprise the kernel:<BR>
</P>
<P>
<B>main.c</B> Initialization and cleanup.</P>
<P>
<B>gp.c</B> The main evaluate-and-breed cycle, and population
statistics calculation.</P>
<P>
<B>eval.c</B> The tree evaluator.</P>
<P>
<B>tree.c </B>Utility routines dealing with trees--counting nodes
and depth, printing, generating random trees, finding
subtrees, copying trees.</P>
<P>
<B>change.c</B> Breeding of the new population each generation.
</P>
<P>
<B>crossovr.c </B>The crossover operator.</P>
<P>
<B>reproduc.c </B>The reproduction operator.</P>
<P>
<B>mutate.c</B> The mutation operator.</P>
<P>
<B>select.c</B> Utility routines for selection methods. </P>
<P>
<B>tournmnt.c</B> The <B>tournament</B> selection method.</P>
<P>
<B>bstworst.c</B> The <B>best,</B> <B>worst</B>, and <B>random</B>
selection methods.</P>
<P>
<B>fitness.c</B> The <B>fitness, fitness_overselect</B>, and <B>inverse_fitness</B>
selection methods.</P>
<P>
<B>genspace.c</B> Utility routines for allocating space to grow
new trees in.</P>
<P>
<B>exch.c</B> The subpopulation exchange system.</P>
<P>
<B>populate.c </B>Utility routines for population--copying, freeing,
random generation.</P>
<P>
<B>ephem.c</B> Utility routines for ephemeral random constants.
</P>
<P>
<B>ckpoint.c</B> Reading and writing checkpoint files. </P>
<P>
<B>event.c </B>System-dependent module for tracking execution
time.</P>
<P>
<B>pretty.c</B> The tree pretty-printer.</P>
<P>
<B>individ.c</B> Utility routines for whole individuals--printing
and calculating size.</P>
<P>
<B>params.c</B> The parameter database.</P>
<P>
<B>random.c</B> The portable pseudorandom number generator, adapted
from <I>Numerical Recipes in FORTRAN</I>.</P>
<P>
<B>memory.c</B> The implementations of <B>MALLOC(), FREE(),</B>
and <B>REALLOC()</B> that track memory usage.</P>
<P>
<B>output.c</B> The output subsystem (<B>oprintf()</B> and <B>oputs(),</B>
among others).</P>
<P>&nbsp;</P>
<P>
<B><A NAME="7.1">7.1 Tree Representation</A></B></P>
<P>
A tree is stored as an array of type <B>lnode.</B> An lnode is
a union which can can a pointer to a function structure, a pointer
to an ephemeral random constant (ERC) structure, or an integer.
The tree is stored in prefix order. The first lnode is always
a pointer to a function. If the function is an ERC terminal, then
the next lnode in the array has the pointer to the ERC structure.
If a function is of type EXPR (the user code controls evaluation
of the function's arguments) then there is an extra lnode just
before the start of each child_it contains an integer, the number
of lnodes used in representing the child subtree. The evaluation
code uses this value to skip the child during evaluation.</P>
<P>
Consider this expression in the symbolic regression problem:</P>
<PRE>
<B>(+ X (iflte X (* .34 X) .56))<BR></B></PRE>
<P>
This would be represented in lil-gp as the array:<FONT SIZE="2"></FONT></P>
<P><FONT SIZE="2">
<BR>
</FONT>
</P>
<TABLE BORDER="0" ALIGN="DEFAULT" WIDTH="100%">
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">+ </TD>
<TD ALIGN="LEFT" VALIGN="TOP">pointer to structure for function +
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">X </TD>
<TD ALIGN="LEFT" VALIGN="TOP">pointer to structure for function X
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">iflte</TD>
<TD>pointer to structure for function iflte
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">1</TD>
<TD>
first argument to iflte takes 1 lnode to store
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">X</TD>
<TD>
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">4 </TD>
<TD>second argument to iflte takes 4 lnodes to store
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">* </TD>
<TD>pointer to structure for function *
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">R</TD>
<TD>pointer to structure for ERC function
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">.34</TD>
<TD>pointer to structure containing ERC value
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">X</TD>
<TD></TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">2</TD>
<TD> third argument to iflte takes 2 lnodes to store
</TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">R </TD>
<TD></TD>
</TR>
<TR>
<TD ALIGN="LEFT" VALIGN="TOP" WIDTH="10%">.56 </TD>
<TD></TD>
</TR>
</TABLE>
<P>
<FONT SIZE="2"><BR>
</FONT>
</P>
<P>
This representation, while it may seem cumbersome, has two major
advantages over a more traditional C representation (with individual
structs for each node, linked by pointers). First, it uses much
less memory--approximately 1-2 words per node versus the 4-5 it
would take otherwise. This is because the structure of the tree
is represented implicitly in the ordering of the nodes rather
than explicitly via pointers. Second, it results in much faster
tree evaluation. For instance, consider the crossover operator.
In the traditional representation crossover is performed by just
swapping two pointers. While this is very fast and easy, over
time it means that the nodes of a given tree become spread out
over the process's address space. On a system with virtual memory,
this slows evaluation (or any traversal of the tree) to a crawl
as the tree is spread across dozens of pages which must constantly
be swapped in and out. lil-gp's representation complicates crossover
somewhat, but leaves each offspring tree as a single continuous
block of memory, able to fit on just one or two pages.<BR>
</P>
<P>
<B><FONT SIZE="4"><A NAME="7.2">7.2 Selection Methods</A></FONT></B></P>
<P>
A selection method is implemented with two functions: one to perform
initialization and cleanup, and another to do the actual selection.
The first function, when called for initialization, creates and
returns a data structure called a selection context. This contains
any state information needed for the selection method. It should
also store a pointer to the population that the selection is being
done on. This structure will be passed to the second function,
which should return an index of an individual within the population.
</P>
<P>
To create a new selection method, it is suggested that you copy
and modify an existing one (the <B>random</B> method in <B>bstworst.c</B>
is an especially simple one). If your of selection can be expressed
as randomly selecting an individual from a set where each individual
has a fixed probability of selection, then there is already an
efficient implementation of the second function. See the code
for the <B>fitness, fitness_overselect,</B> and <B>inverse_fitness</B>
methods for examples</P>
<P>
Finally, you must add a stt_record describing your selection method
to the array <B>select_method_table</B> at the top of <B>select.c.</B>
This lists the names and initialization functions of each selection
method available.<BR>
</P>
<P>
<B><FONT SIZE="4"><A NAME="7.3">7.3 Operator</A></FONT></B></P>
<P>
The breeding of the population is controlled by a table of breeding
phases. This is built from the parameter database at the start
of the run (and whenever <B>rebuild_breeding_table()</B> is called
from user code. Each subpopulation has its own breeding table.
For each phase, there is a stt_record in the table. Each stt_record has
pointers to the four methods for the operator, the rate for that
phase, and a pointer to an operator-specific structure</P>
<P>
To implement a new operator, it is suggested that you copy and
modify the code of an existing operator. The reproduction operator
is the simplest of the three included. Suppose that your new operator
is called &quot;foo&quot;. You would need to provide five functions
(the following naming scheme is strongly recommended):<BR>
</P>
<P>
<B>operator_foo_init</B> Parses the operator's options string
and builds the operator table entry. The kernel
functions <B>parse_o_rama()</B> and <B>free_o_rama()</B> are available
to parse the options string just like the built-in operators do.
<BR>
</P>
<P>
<B>operator_foo_free</B> Frees the operator-specific part of the
operator table entry.<BR>
</P>
<P>
<B>operator_foo_start</B> This is called at the start of breeding.
Selection methods should be initialized here.<BR>
</P>
<P>
<B>operator_foo_end</B> This is called at the end of breeding.
Selection contexts should be freed here.<BR>
</P>
<P>
<B>operator_foo</B> This performs the actual operation.<FONT SIZE="2">
<BR>
</FONT>
</P>
<P>
You then add a stt_record to the array <B>operator_table</B> at the
top of <B>change.c</B>, listing the name of the operator (a string)
and the operator's initialization method (a function pointer),
for example, after adding the foo operator the table would look
like:</P>
<PRE>
<B>operator operator_table[] =</B>
<B>{ { &quot;crossover&quot;, operator_crossover_init },</B>
<B> { &quot;reproduction&quot;, operator_reproduce_init },</B>
<B> { &quot;mutation&quot;, operator_mutate_init },</B>
<B> { &quot;foo&quot;, operator_foo_init },</B>
<B> { NULL, NULL } };<BR>
</B>
</PRE>
<P>
The <B>next</B> field of the new population structure gives the
index at which the operator should place the new individual. After
adding an individual, increment the <B>next</B> field. If your
operator produces multiple offspring with a single call then you
must make sure that you don't overfill the population (the <B>next</B>
field should not exceed the <B>size</B> field). The crossover
operator, for instance, normally produces two offspring on each
call. If it is called when there is only one more space in the
population, it fills the space and throws the other offspring
away.</P>
<P>
Your operator should always add at least one individual to the
new population per call. If it does not, infinite loops may occur
when the <B>probabilistic_operators</B> parameter is off.<BR>
</P>
<P>
<B><FONT SIZE="4"><A NAME="7.4">7.4 Miscellany</A></FONT></B></P>
<P>
This section gives a general overview of how some of the nonobvious
parts of lil-gp work and how they fit together.<BR>
</P>
<P>
<B><A NAME="7.4.1">7.4.1 Tree Generation Spaces</A></B></P>
<P>
When building a new tree, lilgp needs a continuous block of memory
to put the tree in, but can't allocate the final location of the
tree because the size isn't known ahead of time. Therefore, special
blocks of memory are allocated to grow the trees in. Every time
an lnode is added to a tree-in- progress, the function <B>gensp_next</B>
(or <B>gensp_next_int</B>) is called to enlarge the memory block
if needed. Once the tree is finished and its size is known, its
final location is allocated and the tree is copied from the generation
space</P>
<P>
Currently there are two generation spaces needed. If you are implementing
an operator or some- thing that requires three or more trees to
be grown simultaneously, increase <B>GENSPACE_COUNT</B> in <B>defines.h.
</B></P>
<P>
Generation spaces are initially allocated to hold <B>GENSPACE_START</B>
lnodes, and grow in steps of <B>GENSPACE_GROW</B> lnodes. These
constants are in <B>defines.h</B>.<BR>
</P>
<P>
<B><A NAME="7.4.2">7.4.2 Saved Individuals</A></B></P>
<P>
lilgp tracks the best <I>n</I> individuals of each population,
where <I>n</I> is a user-settable parameter. Pointers to these
best individuals are passed to the user function <B>app_end_of_evaluation()</B>.
To ensure that these pointers are always good, the kernel makes
a copy of each individual in the top n and passes the address
of the copy (since the original individual may not survive the
breeding process)</P>
<P>
An individual can be referenced in multiple top-n lists (best-of-gen,
best-of-run, etc.). To avoid making multiple copies, a reference
count is kept for each individual. All these saved individuals
are kept in a linked list, and once per generation a garbage collection
procedure traverses the list and frees any individuals which have
no references to them.<BR>
</P>
<P>
<A NAME="7.4.3"><B>7.4.3 Ephemeral Random Constants</B></A><B></B></P>
<P>
Whenever a new ephemeral random constant terminal is inserted
into a tree (during the generation of the initial population,
or during mutation operations) the function <B>new_ephemeral_const()</B>
is called to create the constant. It calls the user-supplied generation
function to create a new value. Each ERC stt_record stores the value,
along with a reference count of how many tree nodes point to that
value. The ERC records are maintained in a linked list. Once per
generation, a garbage collection routine traverses the linked
list and removes any ERCs which are no longer needed (that have
a reference count of zero)</P>
<P>
The ERC records are not allocated individually but in large blocks.
This ensures that all the ERCs are kept on a few memory pages
at most, which reduces the need for paging during evaluation (when
there are many scattered references to the ERCs). When an ERC
is freed by the garbage collection routine, it is added to the
end of the free list. If the free list ever becomes empty, a new
large block of ERC records is allocated and all of them added
on to the end of the free list</P>
<P>
Pointers to the blocks themselves are kept in an array so that
the blocks can be freed at the conclusion of the run. Since the
number of blocks can increase, this array is reallocated as necessary.
<BR>
</P>
</BODY>
</HTML>