We make science discovery happen

This is the page dedicated to help users of data mining web application (the beta release currently available here), in case of selection of GSOM (General purpose SOM) 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.

There are 3 sub-sections, related with the use cases that users can select to perform experiments, respectively Train, Test and Run.

- Input Dataset file
- Number of clusters/neurons (I layer)
- Number of Iterations
- diameter
- Number of Layers
- Initial variance (I layer)
- Final variance (I layer)
- Initial learning rate (I layer)
- Final learning rate (I layer)
- Number of clusters/neurons (II Layer)
- Initial variance (II Layer)
- Final variance (II Layer)
- Initial learning rate (II Layer)
- Final learning rate (II Layer)
- Number of clusters/neurons (III layer)
- Initial variance (III layer)
- Final variance (III layer)
- Initial learning rate (III layer)
- Final learning rate (III layer)

**this parameter is a required field!**

Input dataset file.

**this parameter is a required field!**

Number of clusters/neurons (first layer), (always integer equal to or greater than 2).

Note that number of neurons corresponds to the number of searched clusters

If left empty, its default is 2

Number of training iterations.

If left empty, its default is 1000

pixel neighborhood, (always integer greater than 2).

If left empty, its default is 3

Number of layers in case of choice of hierarchical Multi layer clustering model, (always integer between 1 and 3).

If left empty, its default is 1 (i.e. no multi layer clustering)

Initial variance (first layer), Real number greater than 0 used on neighborhood function.

If left empty, its default is 3.0

Final variance (first layer), Real number between 0 and value of initial variance of the first layer.

If left empty, its default is 0.001

Initial learning rate (first layer), Real number between 0 and 1.

If left empty, its default is 0.9

Final learning rate (first layer), Real number between 0 and 1 (including 0) STRICTLY LESS than value of initial learning rate of the first layer.

If left empty, its default is 0.001

Number of clusters/Neurons (second Layer). Only in case of Multi layer clustering

Integer number greater than 2 less than number of neurons of first layer and greater than number of neurons of third layer (if specified).

If left empty, its default is 30

Initial variance (second Layer), Real number greater than 0.

If left empty, its default is 3.0

Final variance (second Layer), Real number between 0 and value of initial variance of the second layer.

If left empty, its default is 0.001

Initial learning rate (second Layer), Real number between 0 and 1.

If left empty, its default is 0.9

Final learning Rate (second Layer), Real number between 0 and 1 (including 0) STRICTLY LESS than value of initial learning rate of the second layer.

If left empty, its default is 0.001

Number of clusters/neurons (third layer), Integer number greater than 2 less than number of neurons of second layer.

If left empty, its default is 20

Initial variance (third layer), Real number greater than 0.

If left empty, its default is 3.0

Final variance (third layer), Real number between 0 and value of initial variance of the third layer.

If left empty, its default is 0.001

Initial learning Rate (third layer), Real number between 0 and 1.

If left empty, its default is 0.9

Final learning Rate (third layer) Real number between 0 and 1 (including 0) STRICTLY LESS than value of initial learning rate of the third layer.

If left empty, its default is 0.001

(See the user manual for more details)

- Input Dataset file
- Configuration file for the net
- epochsNumber
- diameter
- NumberLayers
- splitPercentage

**this parameter is a required field!**

Input dataset file.

**this parameter is a required field!**

Configuration file of net given as output from a previous USE CASE TRAIN".

Number of training iterations.

If left empty, its default is 50

pixel neighborhood, (always integer greater than 2).

If left empty, its default is 3

Number of layers in case of choice of hierarchical Multi layer clustering model, (always integer between 1 and 3).

If left empty, its default is 1

Splitting percentage of input dataset, Real number between 0 and 100.

If left empty, its default is 65%

(See the user manual for more details)

- Input Dataset file
- Configuration file for the net
- epochsNumber
- diameter
- NumberLayers

**this parameter is a required field!**

Input dataset file (FITS IMAGE).

**this parameter is a required field!**

Configuration file of net given as output from a previous USE CASE TRAIN".

Number of training iterations.

If left empty, its default is 50

pixel neighborhood, (always integer greater than 2).

If left empty, its default is 3

Number of layers in case of choice of hierarchical Multi layer clustering model, (always integer between 1 and 3).

If left empty, its default is 1

(See the user manual for more details)