Logistic Regression Classification

Author(s)	O.Conrad (c) 2019
Library ID	imagery_opencv
Tool ID	12
Version	1.0
Menu	Imagery \| Classification \| Machine Learning

Description

Integration of the OpenCV Machine Learning library for Logistic Regression based classification of gridded features. Optimization algorithms like Batch Gradient Descent and Mini-Batch Gradient Descent are supported in Logistic Regression. It is important that we mention the number of iterations these optimization algorithms have to run. The number of iterations can be thought as number of steps taken and learning rate specifies if it is a long step or a short step. This and previous parameter define how fast we arrive at a possible solution. In order to compensate for overfitting regularization can be performed. (L1 or L2 norm). Logistic regression implementation provides a choice of two training methods with Batch Gradient Descent or the Mini-Batch Gradient Descent.

References

OpenCV - Open Source Computer Vision

OpenCV - Machine Learning Overview

Parameters

	Name	Type	Identifier	Description	Constraints
Input	Features	grid list, input	`FEATURES`	-	-
	Training Samples	table, input	`TRAIN_SAMPLES`	Provide a class identifier in the first field followed by sample data corresponding to the input feature grids.	-
	Training Areas	shapes, input	`TRAIN_AREAS`	-	-
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	Normalize	boolean	`NORMALIZE`	-	Default: 0
	Colors from Features	boolean [GUI]	`RGB_COLORS`	Use the first three features in list to obtain blue, green, red components for class colour in look-up table.	Default: 0
	Grid System	grid system	`GRID_SYSTEM`	-	-
	Training	choice	`MODEL_TRAIN`	-	Available Choices: [0] training areas [1] training samples [2] load from file Default: 0
	Class Identifier	table field	`TRAIN_CLASS`	-	-
	Buffer Size	floating point number	`TRAIN_BUFFER`	For non-polygon type training areas, creates a buffer with a diameter of specified size.	Minimum: 0.000000 Default: 1.000000
	Load Model	file path	`MODEL_LOAD`	Use a model previously stored to file.	-
	Save Model	file path	`MODEL_SAVE`	Stores model to file to be used for subsequent classifications instead of training areas.	-
	Learning Rate	floating point number	`LOGR_LEARNING_RATE`	The learning rate determines how fast we approach the solution.	Minimum: 0.000000 Default: 1.000000
	Number of Iterations	integer number	`LOGR_ITERATIONS`	-	Minimum: 1 Default: 300
	Regularization	choice	`LOGR_REGULARIZATION`	-	Available Choices: [0] disabled [1] L1 norm [2] L2 norm Default: 0
	Training Method	choice	`LOGR_TRAIN_METHOD`	-	Available Choices: [0] Batch Gradient Descent [1] Mini-Batch Gradient Descent Default: 0
	Mini-Batch Size	integer number	`LOGR_MINIBATCH_SIZE`	-	Minimum: 1 Default: 1

Command Line


Usage: saga_cmd imagery_opencv 12 [-FEATURES ] [-NORMALIZE ] [-CLASSES ] [-CLASSES_LUT ] [-MODEL_TRAIN ] [-TRAIN_SAMPLES ] [-TRAIN_AREAS ] [-TRAIN_CLASS ] [-TRAIN_BUFFER ] [-MODEL_LOAD ] [-MODEL_SAVE ] [-LOGR_LEARNING_RATE ] [-LOGR_ITERATIONS ] [-LOGR_REGULARIZATION ] [-LOGR_TRAIN_METHOD ] [-LOGR_MINIBATCH_SIZE ]
  -FEATURES:                Features
	grid list, input
  -NORMALIZE:               Normalize
	boolean
	Default: 0
  -CLASSES:                 Classification
	grid, output
  -CLASSES_LUT:             Look-up Table
	table, output, optional
  -MODEL_TRAIN:
                                 Training
                                 	choice
                                 	Available Choices:
                                 	[0] training areas
                                 	[1] training samples
                                 	[2] load from file
                                 	Default: 0
  -TRAIN_SAMPLES:           Training Samples
	table, input
  -TRAIN_AREAS:             Training Areas
	shapes, input
  -TRAIN_CLASS:             Class Identifier
	table field
  -TRAIN_BUFFER:         Buffer Size
	floating point number
	Minimum: 0.000000
	Default: 1.000000
  -MODEL_LOAD:              Load Model
	file path
  -MODEL_SAVE:              Save Model
	file path
  -LOGR_LEARNING_RATE:   Learning Rate
	floating point number
	Minimum: 0.000000
	Default: 1.000000
  -LOGR_ITERATIONS:         Number of Iterations
	integer number
	Minimum: 1
	Default: 300
  -LOGR_REGULARIZATION:
                                 Regularization
                                 	choice
                                 	Available Choices:
                                 	[0] disabled
                                 	[1] L1 norm
                                 	[2] L2 norm
                                 	Default: 0
  -LOGR_TRAIN_METHOD:
                                 Training Method
                                 	choice
                                 	Available Choices:
                                 	[0] Batch Gradient Descent
                                 	[1] Mini-Batch Gradient Descent
                                 	Default: 0
  -LOGR_MINIBATCH_SIZE:     Mini-Batch Size
	integer number
	Minimum: 1
	Default: 1