GRASS logo

NAME

i.pr.model - Module to generate model from features file. i.pr: Pattern Recognition environment for image processing. Includes kNN, Decision Tree and SVM classification techniques. Also includes cross-validation and bagging methods for model validation.

KEYWORDS

imagery, image processing, pattern recognition

SYNOPSIS

i.pr.model
i.pr.model help
i.pr.model [-gtsn] features=string model=string [validation=string] [test=string] [npc=integer] [bagging=integer] [boosting=integer] [reg=integer] [misclass_ratio=float] [reg_verbose=integer] [progressive_error=integer] [cost_boosting=float] [weights_boosting=integer] [parallel_boosting=integer] [soft_margin_boosting=integer] [tree_stamps=integer] [tree_minsize=integer] [tree_costs=integer[,integer,...]] [svm_kernel=string] [svm_kp=float] [svm_C=float] [svm_cost=float] [svm_tol=float] [svm_eps=float] [svm_l1o=integer] [svm_maxloops=integer] [svm_verbose=integer] [nn_k=integer] [--verbose] [--quiet]

Flags:

-g
selected model: gaussian mixture.
-t
selected model: classification trees.
-s
selected model: support vector machines.
-n
selected model: nearest neighbor.
--verbose
Verbose module output
--quiet
Quiet module output

Parameters:

features=string
Input file containing the features (output of i.pr_features).
model=string
Name of the output file containing the model.
validation=string
Input file containing other features for the training
and for the evaluation of the performances of the model
on an independent test set (for regularized AdaBoost).
test=string
Input file containing other features for the evaluation of the
performances of the model on an independent test set.
npc=integer
Number of the principal components to be used for the model development.
If not set all the principal components will be used.
Ignored if features does not contain principal components model.
bagging=integer
Number of bagging models.
If bagging = 0 the classical model will be implemented.
Implemented for trees and svm only.
Default: 0
boosting=integer
Number of boosting models.
If boosting = 0 the classical model will be implemented.
Implemented for trees and svm only.
Default: 0
reg=integer
Number of misclassification ratio intervals.
Default: 0
misclass_ratio=float
For regularized AdaBoost: misclassification ratio for
hard point shaving and compute the new model.
Default: 1.00
reg_verbose=integer
For regularized AdaBoost:
- if it is set = 1 the current value of misclassification
ratio and the current error will be printed.
- if it is set to >1 the number of
loops, accuracy and the current value of misclassification ratio
will be printed.
For shaving and compute:
- if it is set >0 the numbers of samples shaved will be printed.
Default: 0
progressive_error=integer
Progressive estimate of the model error
increasing the number of aggregated models
Options: 0,1
Default: 0
cost_boosting=float
Cost parameter for the implementation of cost-sensitive procedure(w in [0,2]).
w>1 results higher weight on examples of class 1.
w<1 results higher weight on examples of class -1.
w=1 corresponds to standard Adaboost.
Default: 1.0
weights_boosting=integer
For boosting: if weights_boosting = 1, a file containing the evolution
of the weights associated to data points will be produced.
Options: 0,1
Default: 0
parallel_boosting=integer
For boosting: number of true boosting steps for parallel boosting.
Implemented only for trees!!
Default: 0
soft_margin_boosting=integer
For boosting: if soft_margin_boosting = 1, sof margin of Ababoost
will bee used. Implemented only with trees. (Sperimental!!!!!!!!!!)
Options: 0,1
Default: 0
tree_stamps=integer
For trees: if tree_stamps = 1, a single split tree will be procuded,
if tree_stamps = 0, a classical tree will be procuded.
Options: 0,1
Default: 0
tree_minsize=integer
For trees: minimum number of examples containined
into a node for splitting the node itself
Default: 0
tree_costs=integer[,integer,...]
For trees: misclassification costs for each class
svm_kernel=string
For svm: type of employed kernel.
Options: gaussian,linear,2pbk
Default: linear
svm_kp=float
For svm: kernel parameter (Required parameter if you are using gaussian kernel).
svm_C=float
For svm: optimization parameter (Required parameter).
svm_cost=float
Cost parameter for the implementation of cost-sensitive procedure(w in [-1,1]).
w>0 results higher weight on examples of class 1.
w<0 results higher weight on examples of class -1.
w=0 corresponds to standard SVM.
Not yet implemented (and may be it will be never implemented)
for bagging and boosting
Default: 0.0
svm_tol=float
For svm: tollerance parameter.
Default: 0.001
svm_eps=float
For svm: epsilon.
Default: 0.001
svm_l1o=integer
For svm: leave 1 out error estimate.
Options: 0,1
Default: 0
svm_maxloops=integer
For svm: maximum number of optimization steps.
Default: 1000
svm_verbose=integer
For svm: if it is set to 1 the number of loops will be printed.
Options: 0,1
Default: 0
nn_k=integer
For nn: Number of neighbor to be considered during the test phase.
Default: 1

Main index - imagery index - Full index

© 2003-2016 GRASS Development Team