/* Copyright (c) 1993 by The Johns Hopkins University */



PEBLS: Parallel Exemplar-Based Learning System

For more information, please contact:


	 Steven Salzberg
	 Dept. of Computer Science
	 Johns Hopkins University
	 Baltimore, MD  21210

	 Email:  salzberg@cs.jhu.edu
	 Phone:  (410) 516-8438


The following files are included:



DOCUMENTATION:

	pebls.doc	The Complete PEBLS 2.0 Manual

	/mljournal-paper

		pebls.ps      	Salzberg and Cost, A Weighted Nearest
				Neighbor Algorithm for Learning with
				Symbolic Features,  Machine Learning,
				10:1, 57-78 (1993)


		pebls-fig1.ps 	Figure 1 from the paper
		pebls-fig2.ps 	Figure 2 from the paper
		pebls-fig3.ps 	Figure 3 from the paper
		pebls-fig4.ps 	Figure 4 from the paper



SOURCE CODE:

	pebls.c 	Core routines for training and testing.
	init.c		Initialization procedures.
	readers.c	Data input routines for all formats
	weights.c	Exemplar and feature weighting functions
	metric.c	Distance metric functions
	symtab.c	Symbol table functions
	util.c		Miscellaneous Utility functions
	output.c	Output procedures

	pelbs.h		Data types, defined constants
	config.h	System configuration constants




EXAMPLE DATA SETS

	protein.dat	Zhang's collection of non-homologous
			protein sub-units (113 total)
			(Converted to PEBLS format)

	protein.pcf	Sample configuration file for Zhang's 
			non-homologous protein data
	

	protein_small.dat  
			Small protein set for fast experiments 
			when learning to use PEBLS.
			(Expect poor performance.)

	protein_small.pcf
			Configuration file for small protein set.


	dna.dat		DNA promotor sequence data (106 total)
			53 positive instances
			53 negative instances

	dna.pcf		Configuration file for
			106 DNA promotor sequences


	sunburn.dat	Sunburn data taken from an example
			in Winston's Text: _Artificial Intelligence_

	sunburn.pcf	Sunburn configuration file.



OTHER FILES

	makefile	makefile for compiling PEBLS 2.0
	README		This file!