Probabilistic Principal Surface help page
This is the page dedicated to help users of data mining web application (the beta release currently available here), in case of selection of PPS Model to make experiments. This page is also directly reachable from the web application, in case users select the help button. The following contents are mainly dedicated to assist the user during the model parameter selection and setup, by giving details about each parameter, its role in the model, default value and suggestions about the right choice.
Functionality: Feature Extraction with PPS
Parameter specifications
Use Case: TRAIN
- Input Dataset File
- Target Column Presence
- Enable Central Latent Variable
- Number of Latent Bases
- Elevation
- Rotation
- Clamping Factor
- Regularization
- Tolerance
- Number of Iterations
this parameter is a field required!
This is the dataset file to be used as input. It is typically composed by objects (rows) and their features (columns), where all objects must have same number of columns. Th format of this file must be one of the types allowed by the application (ASCII, FITS-table, CSV, VOTABLE). The format internally used is CSV. Whenever available, if the file contains also a descriptive target column, it must be included as last column.
This optional parameter should be enabled if the input dataset file contains as last column any descriptive information. So far, enable the checkbox if your input file has this kind of last column. Remeber that this column is not used as training parameter, but only as additional information, eventually useful during post-processing statistical analysis.
By default this field is not selected.
Parameter to be selected if user wants to permit the creation of a central latent variable in the spherical manifold.
By default this field is not selected.
This parameter indicates the number of latent basis functions distributed
The allowed value range is an integer from the range ]0, 300[
By default the value is 10
It is one of the coordinates, expressed in degree, which is used to create the latent variable distribution on the spherical manifold.
The formula to calculate the number of latent variables is N=(((180/Elevation)-1(360/Rotation))+2)
The allowed value range is an integer from the range [5, 10, 15, 20, 30, 45, 60, 90]
By default the value is 90
It is one of the coordinates, expressed in degree, which is used to create the latent variable distribution on the spherical manifold. (See also the Elevation parameter details).
The allowed value range is an integer from the range [5, 10, 15, 20, 30, 45, 60, 90]
By default the value is 90
This parameter determines the orientation of the covariance matrix obtained from the gaussian mixture, used to measure the latent variable centroid.
The allowed value range is a real from the range ]0, 1[
By default the value is 0.5
This parameter performs a regularization of the responsibility matrix, by enhancing the weight update.
The allowed value range is a real < 20
By default the value is 1.1
This parameter is one of the two stopping criteria, based on the maximum likelihood of the output clusters
This parameter is one of the two stopping criteria. Use carefully this value, because it can mostly affect the execution time of the experiment.
The allowed value range is an integer in the range ]0, 1000[
By default the value is 10
Copyright © 2010 DAME Collaboration. All right reserved.