patlrn program

• By downloading this code you agree to the Source Code Use License (PDF).

• Pascal source code: patlrn.p (wget instructions)
• Instructions on compiling
• MacOS binary: patlrn
• Alphabetic List of Delila Programs
• Delila Programs by Most Recent Update
• Please report broken links
• delilabundle.zip = All Programs and MacOS Binaries
• Copyright Statement for Delila Programs

Documentation for the patlrn program is below, with links to related programs in the "see also" section.

{   version = 3.32; (* of patlrn.p 2016 Jan 24}

(* begin module describe.patlrn *)
(*
name
      patlrn: pattern learning

synopsis
      patlrn(funcbook: in, funcinst: in, nfuncbook: in, nfuncinst: in,
            pattern: out, start: in, minmax: in, ignore: in, patlrnp: in,
            output: out)

files
      funcbook: the book of sequences belonging to the functional class;
      funcinst: the instructions for funcbook, for aligning the sequences;
      nfuncbook: the book of sequences for the nonfunctional class;
      nfuncinst: the instructions for nfuncbook, for aligning the seqs;
      pattern: the resulting wmatrix which separates the classes;
      start: a matrix for initializing wmatrix to. it is initialized to
            all 0's if this file is empty;
      minmax: to set the values of funcmin (the minimum value for a functional
            sequence) and nfuncmax (the maximum value for a nonfunctional
            sequence). if this file is empty they are set to 1 and 0,
            respectively, and vary along with the matrix;
      ignore: a file specifying regions of the sequences which are to be
            ignored in the learning process; the maximum number of regions
            which can be ignored is set by the constant 'maxignore';  the
            file must contain two integers per line, the first specifying
            the 5' end and the second the 3' end of the region to be ignored.
      patlrnp: parameter file for setting maxtimes, the number of times
            through all the sequences before stopping without a solution;
      output: for messages to the user.

description
      patlrn uses the 'perceptron' algorithm to find a weighting function
      (a 'wmatrix') which serves to distinguish the sequences in the two
      classes from one another.  our paper, stormo et.al., nar 10, 2995 (1982),
      describes the algorithm in detail and gives an example of its use.

see also
      patlst.p, patana.p, patser.p, patval.p, sepa.p

author
      gary d. stormo (modified by tom schneider)

bugs
      the section of code for ignoring regions of the sequences in the
      learning process (i.e., when the file 'ignore' is not empty) has
      been overlayed over the rest of the code, rather than worked into
      it, and consequently, using this feature can be quite inefficient.

technical note
      the program will be more efficient if the constant 'dnamax' in the
      module 'book.const' is made to be the size of the sequences used
      by the program.  for instance, setting it to whatever 'maxmatrix'
      is would be a good idea.
*)
(* end module describe.patlrn *)
{This manual page was created by makman 1.45}