LibOPF Incremental is a library based on LibOPF with an incremental
version of OPF Classifier.

This code is related to the paper "An incremental linear-time learning algorithm for the
Optimum-Path Forest classifier", submitted to Pattern Recognition Letters.

Moacir Ponti
Mateus Riva

ICMC / USP 2016

For instruction about the original LibOPF check README_LibOPF.

To run the main experiments described in the paper, run the shell script bin/simexp.sh.

Programs included
=================

In this package, we have the following programs:

opf_incremental_split:

	This is a program to randomly split the dataset into a training and testing set, then further split the training dataset into many smaller ones, to simulate incremental learning.

opf_includenew:

	This is a program to execute the classification of new samples and to add them to the trained model, maintaining the model's properties.

opf_includedift:

	This is a program to execute the classification of new samples and to add them to the trained model as prototypes.


'opf_incremental_split' Usage
=================

Usage: opf_incremental_split <P1> <P2> <P3> <P4> <P5>

	P1: input dataset in the OPF file format
	P2: percentage for the training sets total size [0,1]
	P3: percentage for the test set size [0,1]
	P4: percentage of the initial training set
	P5: total of incremental training sets, including the initial
	P6: normalize features? 1 - Yes  0 - No

The sum P2+P3 must be 1.

The features are normalized with the following equation: 

N_i = (F_i - M_i)/S_i, 

where F_i, M_i and S_i are, respectively, the feature i, the average
of F_i and the standard deviation of F_i in the dataset.

The program splits the dataset into P5+1 new files.

'opf_includenew' Usage
=================

Usage: opf_includenew <P1> <P2>

	P1: new data to include in classifier (also in the OPF file format)
	P2: new training set (also in the OPF file format)


	The program designs a classifier from training.opf and the data in P1 and outputs it in a
	file named P2.opf, which is used by 'opf_classify' for
	testing.

	The opf_includenew also outputs the following files:
	-> .out: it contains the predicted labels (incremental training phase)
	-> .time: it contains the execution time in seconds (incremental training phase)
	-> .acc: it contains the accuracy (incremental training phase)



'opf_includedift' Usage
=================

Usage: opf_includedift <P1> <P2>

	P1: new data to include in classifier (also in the OPF file format)
	P2: new training set (also in the OPF file format)


	The program designs a classifier from training.opf and the data in P1 and outputs it in a
	file named P2.opf, which is used by 'opf_classify' for
	testing.

	The opf_includedift also outputs the following files:
	-> .out: it contains the predicted labels (incremental training phase)
	-> .time: it contains the execution time in seconds (incremental training phase)
	-> .acc: it contains the accuracy (incremental training phase)



Tools included
=================

In this package, we have the following tools:

base_creator:

	This is a program to create the BaseN datasets.

'base_creator' Usage
=================

Usage: base_creator <P1> <P2> <P3>


	P1: name of the dataset
	P2: type of the dataset (1, 2 or 3)
	P3: size of the dataset (in sample count)


	The program creates a synthetic dataset based on predefined rules.