Train Dimensionality Reduction

Brief Description

Train a dimensionality reduction model

Tags

Learning

Long Description

Trainer for dimensionality reduction algorithms (autoencoders, PCA, SOM). All input samples are used to compute the model, like other machine learning models.
The model can be used in the ImageDimensionalityReduction and VectorDimensionalityReduction applications.

Parameters

• Input and output data (io): This group of parameters allows setting input and output data.

  • Input Vector Data (io.vd): Input geometries used for training (note : all geometries from the layer will be used)

  • Output model (io.out): Output file containing the estimated model (.txt format).

  • Input XML image statistics file (io.stats): XML file containing mean and variance of each feature.




• Field names to be used for training. (feat): List of field names in the input vector data used as features for training.

• algorithm to use for the training (algorithm): Choice of the dimensionality reduction algorithm to use for the training.

  • OTB SOM (som): This group of parameters allows setting SOM parameters.

    • Map size (algorithm.som.s): Sizes of the SOM map (one per dimension). For instance, [12;15] means a 2D map of size 12x15. Support2D to 5D maps.

    • Neighborhood sizes (algorithm.som.n): Sizes of the initial neighborhood in the SOM map (one per dimension). The number of sizes should be the same as the map sizes

    • NumberIteration (algorithm.som.ni): Number of iterations for SOM learning

    • BetaInit (algorithm.som.bi): Initial learning coefficient

    • BetaFinal (algorithm.som.bf): Final learning coefficient

    • InitialValue (algorithm.som.iv): Maximum initial neuron weight

  • Shark Autoencoder (autoencoder): This group of parameters allows setting Shark autoencoder parameters.

    • Maximum number of iterations during training (algorithm.autoencoder.nbiter): The maximum number of iterations used during training.

    • Maximum number of iterations during training (algorithm.autoencoder.nbiterfinetuning): The maximum number of iterations used during fine tuning of the whole network.

    • Epsilon (algorithm.autoencoder.epsilon): Epsilon

    • Weight initialization factor (algorithm.autoencoder.initfactor): Parameter that control the weight initialization of the autoencoder

    • Size (algorithm.autoencoder.nbneuron): The number of neurons in each hidden layer.

    • Strength of the regularization (algorithm.autoencoder.regularization): Strength of the L2 regularization used during training

    • Strength of the noise (algorithm.autoencoder.noise): Strength of the noise

    • Sparsity parameter (algorithm.autoencoder.rho): Sparsity parameter

    • Sparsity regularization strength (algorithm.autoencoder.beta): Sparsity regularization strength

    • Learning curve (algorithm.autoencoder.learningcurve): Learning error values

  • Shark PCA (pca): This group of parameters allows setting Shark PCA parameters.

    • Dimension of the output of the pca transformation (algorithm.pca.dim): Dimension of the output of the pca transformation.




• Available RAM (Mb) (ram): Available memory for processing (in MB)

• Load otb application from xml file (inxml): Load otb application from xml file

• Save otb application to xml file (outxml): Save otb application to xml file

Limitations

None

Authors

OTB-Team

See also

ImageDimensionalityReduction, VectorDimensionalityReduction

Example of use

• Values for GUIs:

• io.vd: cuprite_samples.sqlite

• io.out: mode.ae

• algorithm: pca

• algorithm.pca.dim: 8

• feat: value_0 value_1 value_2 value_3 value_4 value_5 value_6 value_7 value_8 value_9

• Command line to execute:

otbcli_TrainDimensionalityReduction -io.vd cuprite_samples.sqlite -io.out mode.ae -algorithm pca -algorithm.pca.dim 8 -feat value_0 value_1 value_2 value_3 value_4 value_5 value_6 value_7 value_8 value_9