SVM Classification

Author: O.Conrad (c) 2012
Menu: Imagery | Classification | Machine Learning

Description

Support Vector Machine (SVM) based classification for grids.

References

Chang, C.-C. & Lin, C.-J. (2011): A library for support vector machines. ACM Transactions on Intelligent Systems and Technology, vol.2/3, p.1-27. LIBSVM Homepage.

Parameters

	Name	Type	Identifier	Description	Constraints
Input	Grids	grid list, input	`GRIDS`	-	-
Input	Training Areas	shapes, input	`ROI`	-	-
Output	Classification	grid, output	`CLASSES`	-	-
Output	Look-up Table	table, output, optional	`CLASSES_LUT`	A reference list of the grid values that have been assigned to the training classes.	-
Options	Grid System	grid system	`PARAMETERS_GRID_SYSTEM`	-	-
	Scaling	choice	`SCALING`	-	Available Choices: [0] none [1] normalize (0-1) [2] standardize Default: 2
	Verbose Messages	boolean	`MESSAGE`	-	Default: 0
	Model Source	choice	`MODEL_SRC`	-	Available Choices: [0] create from training areas [1] restore from file Default: 0
	Restore Model from File	file path	`MODEL_LOAD`	-	-
	Class Identifier	table field	`ROI_ID`	-	-
	Store Model to File	file path	`MODEL_SAVE`	-	-
	SVM Type	choice	`SVM_TYPE`	-	Available Choices: [0] C-SVC [1] nu-SVC [2] one-class SVM [3] epsilon-SVR [4] nu-SVR Default: 0
	Kernel Type	choice	`KERNEL_TYPE`	linear: u'v polynomial: (gammau'v + coef0)^degree radial basis function: exp(-gamma\|u-v\|^2) sigmoid: tanh(gammau'v + coef0)	Available Choices: [0] linear [1] polynomial [2] radial basis function [3] sigmoid Default: 2
	Degree	integer number	`DEGREE`	degree in kernel function	Default: 3
	Gamma	floating point number	`GAMMA`	gamma in kernel function	Default: 0.000000
	coef0	floating point number	`COEF0`	coef0 in kernel function	Default: 0.000000
	C	floating point number	`COST`	parameter C (cost) of C-SVC, epsilon-SVR, and nu-SVR	Default: 1.000000
	nu-SVR	floating point number	`NU`	parameter nu of nu-SVC, one-class SVM, and nu-SVR	Default: 0.500000
	SVR Epsilon	floating point number	`EPS_SVR`	epsilon in loss function of epsilon-SVR	Default: 0.100000
	Cache Size	floating point number	`CACHE_SIZE`	cache memory size in MB	Default: 100.000000
	Epsilon	floating point number	`EPS`	tolerance of termination criterion	Default: 0.001000
	Shrinking	boolean	`SHRINKING`	whether to use the shrinking heuristics	Default: 0
	Probability Estimates	boolean	`PROBABILITY`	whether to train a SVC or SVR model for probability estimates	Default: 0
	Cross Validation	integer number	`CROSSVAL`	n-fold cross validation: n must > 1	Minimum: 1 Default: 1

Command Line


Usage: saga_cmd imagery_svm 0 [-GRIDS ] [-CLASSES ] [-CLASSES_LUT ] [-SCALING ] [-MESSAGE ] [-MODEL_SRC ] [-MODEL_LOAD ] [-ROI ] [-ROI_ID ] [-MODEL_SAVE ] [-SVM_TYPE ] [-KERNEL_TYPE ] [-DEGREE ] [-GAMMA ] [-COEF0 ] [-COST ] [-NU ] [-EPS_SVR ] [-CACHE_SIZE ] [-EPS ] [-SHRINKING ] [-PROBABILITY ] [-CROSSVAL ]
  -GRIDS:           Grids
	grid list, input
  -CLASSES:         Classification
	grid, output
  -CLASSES_LUT:     Look-up Table
	table, output, optional
  -SCALING:         Scaling
	choice
	Available Choices:
	[0] none
	[1] normalize (0-1)
	[2] standardize
	Default: 2
  -MESSAGE:         Verbose Messages
	boolean
	Default: 0
  -MODEL_SRC:
                         Model Source
                         	choice
                         	Available Choices:
                         	[0] create from training areas
                         	[1] restore from file
                         	Default: 0
  -MODEL_LOAD:      Restore Model from File
	file path
  -ROI:             Training Areas
	shapes, input
  -ROI_ID:          Class Identifier
	table field
  -MODEL_SAVE:      Store Model to File
	file path
  -SVM_TYPE:
                         SVM Type
                         	choice
                         	Available Choices:
                         	[0] C-SVC
                         	[1] nu-SVC
                         	[2] one-class SVM
                         	[3] epsilon-SVR
                         	[4] nu-SVR
                         	Default: 0
  -KERNEL_TYPE:
                         Kernel Type
                         	choice
                         	Available Choices:
                         	[0] linear
                         	[1] polynomial
                         	[2] radial basis function
                         	[3] sigmoid
                         	Default: 2
  -DEGREE:          Degree
	integer number
	Default: 3
  -GAMMA:        Gamma
	floating point number
	Default: 0.000000
  -COEF0:        coef0
	floating point number
	Default: 0.000000
  -COST:         C
	floating point number
	Default: 1.000000
  -NU:           nu-SVR
	floating point number
	Default: 0.500000
  -EPS_SVR:      SVR Epsilon
	floating point number
	Default: 0.100000
  -CACHE_SIZE:   Cache Size
	floating point number
	Default: 100.000000
  -EPS:          Epsilon
	floating point number
	Default: 0.001000
  -SHRINKING:       Shrinking
	boolean
	Default: 0
  -PROBABILITY:     Probability Estimates
	boolean
	Default: 0
  -CROSSVAL:        Cross Validation
	integer number
	Minimum: 1
	Default: 1