Torch Deep Learning

Associated Constructors

Torch Deep Learning

Syntax: Torch Deep Learning(Y( columns ), X( columns ))

Description: Interface to predictive modeling via the Torch Deep Learning add-in

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Columns

Censor

Syntax: obj << Censor( column )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Freq

Syntax: obj << Freq( column )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_freqcol", Numeric, Continuous, Formula( Random Integer( 1, 5 ) ) );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Inputs

Syntax: obj << Inputs( column(s) )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Responses

Syntax: obj << Responses( column(s) )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Subject

Syntax: obj << Subject( column )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Validation

Syntax: obj << Validation( column(s) )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Weight

Syntax: obj << Weight( column )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_weightcol", Numeric, Continuous, Formula( Random Beta( 1, 1 ) ) );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

X

Syntax: obj << X( column(s) )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Y

Syntax: obj << Y( column(s) )

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Item Messages

Change Variables

Syntax: obj << Change Variables

Description: Changes X, Y, and other variables for subsequent models.

JMP Version Added: 19

Compare

Syntax: obj << Compare

Description: Updates the Torch Deep Learning comparison metrics.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Compare( AUC( 1 ) );

Fit

Syntax: obj << Fit

Description: Fits a Torch Deep Learning model. You can specify parameters and fitting specifications within this command.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );

Get Measures

Syntax: obj << Get Measures

JMP Version Added: 19

Redo Analysis

Syntax: obj << Redo Analysis

Description: Rerun this same analysis in a new window. The analysis will be different if the data has changed.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Redo Analysis;

Relaunch Analysis

Syntax: obj << Relaunch Analysis

Description: Return to the launcher for this analysis.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Relaunch Analysis;

Set

Syntax: obj << Set

Description: Specifies parameters for a Torch Deep Learning model.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Set( Epochs( 5 ) ) );

Show Details

Syntax: obj << Show Details( state=0|1 )

Description: Shows more details.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Show Details( 1 ) );

Shared Item Messages

Action

Syntax: obj << Action

Description: All-purpose trapdoor within a platform to insert expressions to evaluate. Temporarily sets the DisplayBox and DataTable contexts to the Platform.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
dt << Bivariate(
    Y( :height ),
    X( :weight ),
    Action( Distribution( Y( :height, :weight ), Histograms Only ) )
);

Apply Preset

Syntax: Apply Preset( preset ); Apply Preset( source, label, <Folder( folder {, folder2, ...} )> )

Description: Apply a previously created preset to the object, updating the options and customizations to match the saved settings.

JMP Version Added: 19

Anonymous preset

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Oneway( Y( :height ), X( :sex ), t Test( 1 ) );
preset = obj << New Preset();
dt2 = Open( "$SAMPLE_DATA/Dogs.jmp" );
obj2 = dt2 << Oneway( Y( :LogHist0 ), X( :drug ) );
Wait( 1 );
obj2 << Apply Preset( preset );

Search by name

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Oneway( Y( :height ), X( :sex ) );
Wait( 1 );
obj << Apply Preset( "Sample Presets", "Compare Distributions" );

Search within folder(s)

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Oneway( Y( :height ), X( :sex ) );
Wait( 1 );
obj << Apply Preset( "Sample Presets", "t-Tests", Folder( "Compare Means" ) );

Copy ByGroup Script

Syntax: obj << Copy ByGroup Script

Description: Create a JSL script to produce this analysis, and put it on the clipboard.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj[1] << Copy ByGroup Script;

Copy Script

Syntax: obj << Copy Script

Description: Create a JSL script to produce this analysis, and put it on the clipboard.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Copy Script;

Get By Levels

Syntax: obj << Get By Levels

Description: Returns an associative array mapping the by group columns to their values.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
biv = dt << Bivariate( X( :height ), Y( :weight ), By( :sex ) );
biv << Get By Levels;

Get ByGroup Script

Syntax: obj << Get ByGroup Script

Description: Creates a script (JSL) to produce this analysis and returns it as an expression.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
t = obj[1] << Get ByGroup Script;
Show( t );

Get Container

Syntax: obj << Get Container

Description: Returns a reference to the container box that holds the content for the object.

JMP Version Added: 19

General

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
t = obj << Get Container;
Show( (t << XPath( "//OutlineBox" )) << Get Title );

Platform with Filter

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
gb = Graph Builder(
    Show Control Panel( 0 ),
    Variables( X( :height ), Y( :weight ) ),
    Elements( Points( X, Y, Legend( 1 ) ), Smoother( X, Y, Legend( 2 ) ) ),
    Local Data Filter(
        Add Filter(
            columns( :age, :sex, :height ),
            Where( :age == {12, 13, 14} ),
            Where( :sex == "F" ),
            Where( :height >= 55 ),
            Display( :age, N Items( 6 ) )
        )
    )
);
New Window( "platform boxes",
    H List Box(
        Outline Box( "Report(platform)", Report( gb ) << Get Picture ),
        Outline Box( "platform << Get Container", (gb << Get Container) << Get Picture )
    )
);

Get Data Table

Syntax: obj << Get Data Table

Description: Returns a reference to the data table.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
t = obj << Get Datatable;
Show( N Rows( t ) );

Get Script

Syntax: obj << Get Script

Description: Creates a script (JSL) to produce this analysis and returns it as an expression.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
t = obj << Get Script;
Show( t );

Get Script With Data Table

Syntax: obj << Get Script With Data Table

Description: Creates a script(JSL) to produce this analysis specifically referencing this data table and returns it as an expression.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
t = obj << Get Script With Data Table;
Show( t );

Get Timing

Syntax: obj << Get Timing

Description: Times the platform launch.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
t = obj << Get Timing;
Show( t );

Get Web Support

Syntax: obj << Get Web Support

Description: Return a number indicating the level of Interactive HTML support for the display object. 1 means some or all elements are supported. 0 means no support.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Bivariate( Y( :Weight ), X( :Height ) );
s = obj << Get Web Support();
Show( s );

Get Where Expr

Syntax: obj << Get Where Expr

Description: Returns the Where expression for the data subset, if the platform was launched with By() or Where(). Otherwise, returns Empty()

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
biv = dt << Bivariate( X( :height ), Y( :weight ), By( :sex ) );
biv2 = dt << Bivariate( X( :height ), Y( :weight ), Where( :age < 14 & :height > 60 ) );
Show( biv[1] << Get Where Expr, biv2 << Get Where Expr );

Ignore Platform Preferences

Syntax: Ignore Platform Preferences( state=0|1 )

Description: Ignores the current settings of the platform's preferences. The message is ignored when sent to the platform after creation.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
dt << Bivariate(
    Ignore Platform Preferences( 1 ),
    Y( :height ),
    X( :weight ),
    Action( Distribution( Y( :height, :weight ), Histograms Only ) )
);

New JSL Preset

Syntax: New JSL Preset( preset )

Description: For testing purposes, create a preset directly from a JSL expression. Like <<New Preset, it will return a Platform Preset that can be applied using <<Apply Preset. But it allows you to specify the full JSL expression for the preset to test outside of normal operation. You will get an Assert on apply if the platform names do not match, but that is expected.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Oneway( Y( :Height ), X( :Age ) );
preset = obj << New JSL Preset( Oneway( Y( :A ), X( :B ), Each Pair( 1 ) ) );
Wait( 1 );
obj << Apply Preset( preset );

New Preset

Syntax: obj = New Preset()

Description: Create an anonymous preset representing the options and customizations applied to the object. This object can be passed to Apply Preset to copy the settings to another object of the same type.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Oneway( Y( :height ), X( :sex ), t Test( 1 ) );
preset = obj << New Preset();

Render Preset

Syntax: Render Preset( preset )

Description: For testing purposes, show the platform rerun script that would be used when applying a platform preset to the platform in the log. No changes are made to the platform.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Oneway( Y( :Height ), X( :Age ) );
obj << Render Preset( Expr( Oneway( Y( :A ), X( :B ), Each Pair( 1 ) ) ) );

Report

Syntax: obj << Report;Report( obj )

Description: Returns a reference to the report object.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
r = obj << Report;
t = r[Outline Box( 1 )] << Get Title;
Show( t );

Save ByGroup Script to Data Table

Syntax: Save ByGroup Script to Data Table( <name>, < <<Append Suffix(0|1)>, < <<Prompt(0|1)>, < <<Replace(0|1)> );

Description: Creates a JSL script to produce this analysis, and save it as a table property in the data table. You can specify a name for the script. The Append Suffix option appends a numeric suffix to the script name, which differentiates the script from an existing script with the same name. The Prompt option prompts the user to specify a script name. The Replace option replaces an existing script with the same name.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj[1] << Save ByGroup Script to Data Table;

Save ByGroup Script to Journal

Syntax: obj << Save ByGroup Script to Journal

Description: Create a JSL script to produce this analysis, and add a Button to the journal containing this script.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj[1] << Save ByGroup Script to Journal;

Save ByGroup Script to Script Window

Syntax: obj << Save ByGroup Script to Script Window

Description: Create a JSL script to produce this analysis, and append it to the current Script text window.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj[1] << Save ByGroup Script to Script Window;

Save Script for All Objects

Syntax: obj << Save Script for All Objects

Description: Creates a script for all report objects in the window and appends it to the current Script window. This option is useful when you have multiple reports in the window.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Save Script for All Objects;

Save Script for All Objects To Data Table

Syntax: obj << Save Script for All Objects To Data Table( <name> )

Description: Saves a script for all report objects to the current data table. This option is useful when you have multiple reports in the window. The script is named after the first platform unless you specify the script name in quotes.

JMP Version Added: 19

Example 1

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj[1] << Save Script for All Objects To Data Table;

Example 2

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
dt << New Column( "_bycol",
    Character,
    Nominal,
    set values( Repeat( {"A", "B"}, N Rows( dt ) )[1 :: N Rows( dt )] )
);
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj[1] << Save Script for All Objects To Data Table( "My Script" );

Save Script to Data Table

Syntax: Save Script to Data Table( <name>, < <<Prompt(0|1)>, < <<Replace(0|1)> );

Description: Create a JSL script to produce this analysis, and save it as a table property in the data table.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Save Script to Data Table( "My Analysis", <<Prompt( 0 ), <<Replace( 0 ) );

Save Script to Journal

Syntax: obj << Save Script to Journal

Description: Create a JSL script to produce this analysis, and add a Button to the journal containing this script.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Save Script to Journal;

Save Script to Report

Syntax: obj << Save Script to Report

Description: Create a JSL script to produce this analysis, and show it in the report itself. Useful to preserve a printed record of what was done.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Save Script to Report;

Save Script to Script Window

Syntax: obj << Save Script to Script Window

Description: Create a JSL script to produce this analysis, and append it to the current Script text window.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Save Script to Script Window;

SendToByGroup

Syntax: SendToByGroup( {":Column == level"}, command );

Description: Sends platform commands or display customization commands to each level of a by-group.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
dt << Distribution(
    By( :Sex ),
    SendToByGroup(
        {:sex == "F"},
        Continuous Distribution( Column( :weight ), Normal Quantile Plot( 1 ) )
    ),
    SendToByGroup( {:sex == "M"}, Continuous Distribution( Column( :weight ) ) )
);

SendToEmbeddedScriptable

Syntax: SendToEmbeddedScriptable( Dispatch( "Outline name", "Element name", command );

Description: SendToEmbeddedScriptable restores settings of embedded scriptable objects.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Reliability/Fan.jmp" );
dt << Life Distribution(
    Y( :Time ),
    Censor( :Censor ),
    Censor Code( 1 ),
    <<Fit Weibull,
    SendToEmbeddedScriptable(
        Dispatch(
            {"Statistics", "Parametric Estimate - Weibull", "Profilers", "Density Profiler"},
            {1, Confidence Intervals( 0 ), Term Value( Time( 6000, Lock( 0 ), Show( 1 ) ) )}
        )
    )
);

SendToReport

Syntax: SendToReport( Dispatch( "Outline name", "Element name", Element type, command );

Description: Send To Report is used in tandem with the Dispatch command to customize the appearance of a report.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
dt << Distribution(
    Nominal Distribution( Column( :age ) ),
    Continuous Distribution( Column( :weight ) ),
    SendToReport( Dispatch( "age", "Distrib Nom Hist", FrameBox, {Frame Size( 178, 318 )} ) )
);

Title

Syntax: obj << Title( "new title" )

Description: Sets the title of the platform.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
obj << Title( "My Platform" );

Top Report

Syntax: obj << Top Report

Description: Returns a reference to the root node in the report.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
obj = Torch Deep Learning( Y( :sex ), X( :picture ), Fit );
r = obj << Top Report;
t = r[Outline Box( 1 )] << Get Title;
Show( t );

View Web XML

Syntax: obj << View Web XML

Description: Returns the XML code that is used to create the interactive HTML report.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class.jmp" );
obj = dt << Bivariate( Y( :Weight ), X( :Height ) );
xml = obj << View Web XML;

Torch Deep Learning Compare

Associated Constructors

Torch Deep Learning Compare

Syntax: Torch Deep Learning Compare

JMP Version Added: 19

Item Messages

AUC

Syntax: obj << AUC( state=0|1 )

Description: Shows or hides the AUROC, which is the area under the receiver operating characteristic curve. On by default.

JMP Version Added: 19

Accuracy

Syntax: obj << Accuracy( state=0|1 )

Description: Shows or hides the accuracy, which is the proportion of correct classifications. On by default.

JMP Version Added: 19

Censor

Syntax: obj << Censor( state=0|1 )

Description: Shows or hides the Censor command On by default.

JMP Version Added: 19

Concordance

Syntax: obj << Concordance( state=0|1 )

Description: Shows or hides the concordance, which is the Harrell C-Index and measures strength of sorting efficiency On by default.

JMP Version Added: 19

Correlation

Syntax: obj << Correlation( state=0|1 )

Description: Shows or hides the Pearson correlation, which is a measure of the strength of the linear relationship. On by default.

JMP Version Added: 19

F1

Syntax: obj << F1( state=0|1 )

Description: Shows or hides the F1 Score, which is the harmonic average of precision and recall. On by default.

JMP Version Added: 19

Freq

Syntax: obj << Freq( state=0|1 )

Description: Shows or hides the Freq column. On by default.

JMP Version Added: 19

H Measure

Syntax: obj << H Measure( state=0|1 )

Description: Shows or hides the H Measure, which measures proportion improvement over baseline. On by default.

JMP Version Added: 19

Hide All Models

Syntax: obj << Hide All Models

Description: Hides all models.

JMP Version Added: 19

LogLoss

Syntax: obj << LogLoss( state=0|1 )

Description: Shows or hides the logarithm of the likelihood-based loss function. On by default.

JMP Version Added: 19

MAE

Syntax: obj << MAE( state=0|1 )

Description: Shows or hides the MAE, which is the mean absolute error. On by default.

JMP Version Added: 19

MCC

Syntax: obj << MCC( state=0|1 )

Description: Shows or hides the Matthews correlation coefficient, which is the Pearson correlation for binary variables. On by default.

JMP Version Added: 19

Misclass

Syntax: obj << Misclass( state=0|1 )

Description: Shows or hides the misclassification rate, which is the proportion of incorrect classifications. On by default.

JMP Version Added: 19

Precision Recall AUC

Syntax: obj << Precision Recall AUC( state=0|1 )

Description: Shows or hides the Precision Recall AUC, which is the area under the precision-recall curve. On by default.

JMP Version Added: 19

Predictors

Syntax: obj << Predictors( state=0|1 )

Description: Shows or hides the Predictors column. On by default.

JMP Version Added: 19

Profit

Syntax: obj << Profit( state=0|1 )

Description: Shows or hides the expected profit. On by default.

JMP Version Added: 19

RMSE

Syntax: obj << RMSE( state=0|1 )

Description: Shows or hides the RMSE, which is the root mean square error. On by default.

JMP Version Added: 19

RSquare

Syntax: obj << RSquare( state=0|1 )

Description: Shows or hides RSquare value, which is the proportion of variability explained. On by default.

JMP Version Added: 19

Remove Hidden Models

Syntax: obj << Remove Hidden Models

Description: Removes all models for which the Show box is not checked.

JMP Version Added: 19

Remove Shown Models

Syntax: obj << Remove Shown Models

Description: Removes all models for which the Show check box is checked and shows the remaining models.

JMP Version Added: 19

Response

Syntax: obj << Response( state=0|1 )

Description: Shows or hides the Response column. On by default.

JMP Version Added: 19

Show All Models

Syntax: obj << Show All Models

Description: Shows all models.

JMP Version Added: 19

Subject

Syntax: obj << Subject( state=0|1 )

Description: Shows or hides the Subject column On by default.

JMP Version Added: 19

Training Metrics

Syntax: obj << Training Metrics( state=0|1 )

Description: Shows or hides all training metrics. On by default.

JMP Version Added: 19

Validation

Syntax: obj << Validation( state=0|1 )

Description: Shows or hides the Validation column. On by default.

JMP Version Added: 19

Validation Metrics

Syntax: obj << Validation Metrics( state=0|1 )

Description: Shows or hides all validation metrics. On by default.

JMP Version Added: 19

Weight

Syntax: obj << Weight( state=0|1 )

Description: Shows or hides the Weight column. On by default.

JMP Version Added: 19

Torch Deep Learning Fit > Post

Item Messages

Actual by Predicted Plots

Syntax: obj << Actual by Predicted Plots( state=0|1 )

Description: Shows or hides a plot using the training data with the predicted values on the X axis and actual values on the Y axis. On by default.

JMP Version Added: 19

Confusion Matrices

Syntax: obj << ( fit[number] << Confusion Matrices( state=0|1 ) )

Description: Shows or hides a crosstabulation matrix of actual and predicted levels. On by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Iris.jmp" );
obj = dt << Torch Deep Learning(
    Y( :Species ),
    X( :Sepal length, :Sepal width, :Petal length, :Petal width ),
    Fit
);
obj << (fit[1] << Confusion Matrices( 1 ));

Contour Profiler.

Syntax: obj << Contour Profiler.

Description: Shows or hides interactive graphs of cross-sections of the prediction function.

JMP Version Added: 19

Decision Thresholds

Syntax: obj << Decision Thresholds( state=0|1 )

Description: Shows or hides decision threshold graphs and tables. On by default.

JMP Version Added: 19

Fit Details

Syntax: obj << Fit Details( state=0|1 )

Description: Shows or hides the statistics for the fitted model. On by default.

JMP Version Added: 19

Lift Curves

Syntax: obj << Lift Curves( state=0|1 )

Description: Plots how much more saturated the top x-percent of predicted values are compared to the whole population.

JMP Version Added: 19

Model Details

Syntax: obj << Model Details( state=0|1 )

Description: Shows or hides model details On by default.

JMP Version Added: 19

Precision Recall Curves

Syntax: obj << Precision Recall Curves( state=0|1 )

Description: Plots the trade-off between precision and recall for different classification thresholds. It is preferred in scenarios where class imbalances exist.

JMP Version Added: 19

Profiler

Syntax: obj << Profiler

Description: Shows or hides the Prediction Profiler.

JMP Version Added: 19

ROC Curves

Syntax: obj << ROC Curves( state=0|1 )

Description: Plots the response-category sorting efficiency of the model predictions.

JMP Version Added: 19

Surface Profiler

Syntax: obj << Surface Profiler

Description: Shows or hides interactive graphs of cross-sections of the prediction function.

JMP Version Added: 19

Torch Deep Learning Fit

Associated Constructors

Post

Syntax: Post

JMP Version Added: 19

Torch Deep Learning Fit

Syntax: Torch Deep Learning Fit

JMP Version Added: 19

Item Messages

Activation

Syntax: obj << Activation( "CELU"|"ELU"|"GELU"|"Hardshrink"|"Hardtanh"|"LeakyReLU"|"LogSigmoid"|"Mish"|"PReLU"|"ReLU"|"ReLU6"|"RReLU"|"SELU"|"Sigmoid"|"SiLU"|"Softplus"|"Softshrink"|"Softsign"|"Tanh"|"Tanhshrink"|"None"="ReLU" )

Description: Specifies the activation function to use after each layer. "ReLU" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Activation( "ReLU" ) ) );

Activations

Syntax: obj << Activations( text )

Description: Specifies a space-delimited list of activation functions to use in sequential layers. This parameter overrides Activation when it is specified, and the last value carries forward.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Activations( "ReLU" ) ) );

Anchor Scale

Syntax: obj << Anchor Scale( number=16 )

Description: Specifies a multiplier applied to an internal range of anchor sizes. Larger values tend to work better for larger boxes. "16" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Anchor Scale( "16" ) ) );

Aspect Sigma

Syntax: obj << Aspect Sigma( number=0 )

Description: Standard deviation of Gaussian aspect ratio deformation "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Aspect Sigma( 0.0 ) ) );

Attention Heads

Syntax: obj << Attention Heads( text=4 )

Description: For transformer models, specifies the number of attention heads as a space delimited list of positive integers, each of which must evenly divide its corresponding layer size. Last value carries forward if necessary. "4" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Attention Heads( 1 ) ) );

Base Activation

Syntax: obj << Base Activation( "CELU"|"ELU"|"GELU"|"Hardshrink"|"Hardtanh"|"LeakyReLU"|"LogSigmoid"|"Mish"|"PReLU"|"ReLU"|"ReLU6"|"RReLU"|"SELU"|"Sigmoid"|"SiLU"|"Softplus"|"Softshrink"|"Softsign"|"Tanh"|"Tanhshrink"|"None"="GELU" )

Description: Specifies the base activation function for Kolmogorov Arnold B Splines. "GELU" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Base Activation( "GELU" ) ) );

Basis Function

Syntax: obj << Basis Function( "Gaussian"|"Linear"|"Quadradic"|"InverseQuadradic"|"MultiQuadric"|"InverseMultiQuadric"|"Spline"|"Poisson1"|"Poisson2"|"Matern32"|"Matern52"="Gaussian" )

Description: For Radial Basis Machine models, specify the basis function. "Gaussian" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning(
    Y( :sex ),
    X( :height, :weight ),
    Fit( Tabular Model( "RadialBasisMachine" ), Basis Function( "Gaussian" ) )
);

Batch Size

Syntax: obj << Batch Size( number=128 )

Description: Specifies the number of rows to randomly sample for each training batch and optimization update. Decrease it to save memory and update gradients more frequently; increase it to pass through the data faster and regularize the model more. "128" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Batch Size( 128 ) ) );

Binary Loss

Syntax: obj << Binary Loss( "BCE"|"SM"="BCE" )

Description: Specifies the loss function for binary responses. Choose from Binary Cross Entropy (BCE) or Soft Margin (SM). "BCE" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Binary Loss( "BCE" ) ) );

Blur Max Sigma

Syntax: obj << Blur Max Sigma( number=0 )

Description: Maximum standard deviation of Gaussian blur "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Blur Max Sigma( 1 ) ) );

Class Loss Weight

Syntax: obj << Class Loss Weight( number=4.0 )

Description: Specifies the multiplier for class loss. "4.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Class Loss Weight( 4.0 ) ) );

Confidence Threshold

Syntax: obj << Confidence Threshold( number=0.05 )

Description: Specifies the confidence score threshold for predicted boxes. Boxes with probability score less than this threshold are dropped. "0.05" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Confidence Threshold( 0.05 ) ) );

Continuous Loss

Syntax: obj << Continuous Loss( "MSE"|"L1"|"SmoothL1"|"Huber"|"Poisson"|"Quantile"|"CoxPH"="MSE" )

Description: Specifies the loss function for continuous responses. Choose from Mean Squared Error (MSE), Mean Absolute Error (L1), Smoothed L1 (with margin), Huber (with margin), or Poisson (for count responses). "MSE" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :weight ), X( :picture ), Fit( Continuous Loss( "MSE" ) ) );

Copy Parameters to Launch

Syntax: obj << Copy Parameters to Launch

Description: Copies the parameter values from this model to the model launch section.

JMP Version Added: 19

Covariance Structure

Syntax: obj << Covariance Structure( "DotProduct"|"Gaussian"="DotProduct" )

Description: For mixed models, specify the covariance structure. "DotProduct" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning(
    Y( :sex ),
    X( :height, :weight ),
    Fit( Tabular Model( "MixedModel" ), Covariance Structure( "DotProduct" ) )
);

Data Threads

Syntax: obj << Data Threads( number=4 )

Description: Specifies the number of threads to use to load data into memory. A number near half the number of actual cores is usually near optimal. "4" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Data Threads( 0 ) ) );

Device

Syntax: obj << Device( "auto"|"cpu"|"cuda:0"|"cuda:1"|"cuda:2"|"cuda:3"="auto" )

Description: Specifies the computational device that Torch uses. "auto" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Device( "cpu" ) ) );

Dilations

Syntax: obj << Dilations( text=1 )

Description: For custom convolutional models, specifies the dilations as a space-delimited list of positive integers. Last value carries forward if necessary. "1" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Dilations( "1" ) ) );

Dropout Probs

Syntax: obj << Dropout Probs( text=0.0 )

Description: Specifies the probabilities of dropout to use after each layer as a space-delimited list of decimals between 0 and 1. Last value carries forward if necessary. "0.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Dropout Probs( "0.1" ) ) );

Epochs

Syntax: obj << Epochs( number=20 )

Description: Specifies the number of iterations through the training data to optimize the loss function for each batch and train the model. "20" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Epochs( 100 ) ) );

Factorization Machine Layers

Syntax: obj << Factorization Machine Layers( text=0 )

Description: Specify a space-separated list of 0s and 1s indicating if factorization machine interactions should be added to each linear layer. Last value carries forward. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning(
    Y( :sex ),
    X( :height, :weight ),
    Fit( Factorization Machine Layers( "1" ) )
);

Fit Ys Separately

Syntax: obj << Fit Ys Separately( state=0 )

Description: Check to fit a distinct model for each Y variable, and uncheck to model them jointly. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex, :height ), X( :picture ), Fit( Model Ys Separately( 1 ) ) );

Fixed Effects

Syntax: obj << Fixed Effects( number=0 )

Description: Specify the number of fixed effects, all of which must be at the beginning of the X variable list "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Fixed Effects( 0 ) ) );

Folder

Syntax: obj << Folder( text )

Description: Select a folder in which to save modeling results. A subfolder for each model is created in this folder.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Folder( "" ) ) );

Frozen Epochs

Syntax: obj << Frozen Epochs( number=0 )

Description: Specifies the number of epochs for which pretrained model bodies remain frozen. After this number there is full training gradients for all parameters. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Frozen Epochs( 3 ) ) );

Generate Python Code

Syntax: obj << Generate Python Code

Description: Creates Python code for model deployment.

JMP Version Added: 19

Grid Size

Syntax: obj << Grid Size( number=5 )

Description: For Kolmogorov Arnold B Spline networks, specifies the number of points in the grid for the spline interpolation. "5" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Grid Size( 5 ) ) );

HFlip Prob

Syntax: obj << HFlip Prob( number=0 )

Description: Probability of horizontal flip "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( HFlip Prob( 0.3 ) ) );

Highway Layers

Syntax: obj << Highway Layers( text=0 )

Description: Specify a space-separated list of nonnegative integers specifying the number of highway layers to insert in the network. Last value carries forward. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Highway Layers( "1" ) ) );

Image Model

Syntax: obj << Image Model( ="LeNet5" )

Description: Specifies the image network architecture to use. Models are ordered by size. Smaller models train faster but may not perform as well as larger models. "LeNet5" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Image Model( "LeNet5" ) ) );

Image Size

Syntax: obj << Image Size( number=28 )

Description: Specifies the size of image to use while training. Input images are transformed to this size square; larger images have higher resolution but slower training times. "28" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Image Size( 28 ) ) );

Kernel Sizes

Syntax: obj << Kernel Sizes( text=3 )

Description: For custom convolutional models, specifies the kernel sizes as a space-delimited list of positive integers. Last value carries forward if necessary. "3" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Kernel Sizes( "3" ) ) );

L1 Penalty

Syntax: obj << L1 Penalty( number=0.0 )

Description: Specifies a multiplier for the sum of absolute values of weight parameters to be added to the loss and induce sparsity. "0.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( L1 Penalty( 0.0001 ) ) );

Layer Sizes

Syntax: obj << Layer Sizes( text=16 )

Description: Specifies output sizes of hidden layers as a space-delimited list of integers (actual sizes) or decimals (multipliers of the previous layer size). The final value is the embedding size. "16" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Layer Sizes( "16" ) ) );

Learning Rate

Syntax: obj << Learning Rate( number=0.001 )

Description: Specifies the learning rate. Smaller learning rates tend to fit better but require more iterations to converge, whereas larger learning rates fit faster. "0.001" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Learning Rate( 0.001 ) ) );

Margin

Syntax: obj << Margin( number=1.0 )

Description: Specifies the margin used in margin-based loss functions. Larger values should produce larger embedding distances between nominal responses with different levels, but may adversely affect training. "1.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Margin( 1.0 ) ) );

Max Boxes

Syntax: obj << Max Boxes( number=5 )

Description: Specifies the maximum number of predicted boxes per image. "5" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Max Boxes( 5 ) ) );

Max Seq Length

Syntax: obj << Max Seq Length( number=512 )

Description: For text models, specifies the maximum number of tokens to create for each text item. "512" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Chips.jmp" );
Torch Deep Learning(
    Y( :Buy again? ),
    X( :Potato Chip Product Review ),
    Fit( Max Seq Length( 512 ) )
);

Mixup Portion

Syntax: obj << Mixup Portion( number=0.0 )

Description: Specifies portion of mixup samples to add to each training batch. For example, if Batch Size is 128 and Mixup Portion is 0.5, then 64 mixup samples are added. "0.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Mixup Portion( 0.5 ) ) );

NMS Threshold

Syntax: obj << NMS Threshold( number=0.5 )

Description: Specifies the non-maximum suppression threshold for predicted boxes. Overlapping boxes with IOU values above this threshold are dropped. "0.5" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( NMS Threshold( 0.5 ) ) );

Noise Max Sigma

Syntax: obj << Noise Max Sigma( number=0 )

Description: Maximum standard deviation of additive Gaussian noise "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Noise Max Sigma( 1 ) ) );

Nominal Image Threshold

Syntax: obj << Nominal Image Threshold( number=10 )

Description: Specifies the cutoff for determining if images in a column are nominal or continuous. If the number of unique pixel levels is <= this number, then the images are considered to be nominal. "10" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Nominal Image Threshold( 10 ) ) );

Nominal Loss

Syntax: obj << Nominal Loss( "NLL"="NLL" )

Description: Specifies the loss function for nominal responses. Choose from Negative Loglikelihood (NLL). "NLL" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Nominal Loss( "NLL" ) ) );

Norm

Syntax: obj << Norm( "None"|"Batch"|"Group"|"Instance"="Batch" )

Description: Specifies the type of normalization to apply to each MLP layer. "Batch" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Norm( "Batch" ) ) );

Norm First

Syntax: obj << Norm First( "None"|"Batch"="Batch" )

Description: Specifies the type of normalization to apply to the input data to the tabular model. Batch norm effectively centers and scales each input. "Batch" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Norm First( "Batch" ) ) );

Num Linear

Syntax: obj << Num Linear( number=1 )

Description: For custom convolutional and message passing models, specifies the number of linear layers at the end of Layer Sizes. "1" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Num Linear( 1 ) ) );

Optimizer

Syntax: obj << Optimizer( "Adam"|"AdamW"|"SGD"|"SGDAGC"="AdamW" )

Description: Specifies the optimization method. Choose between Adaptive moment estimation (Adam), Adam weight decay (AdamW), Stochastic Gradient Descent (SGD), or SGD with Adaptive Gradient Clipping (SGDAGC). "AdamW" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Optimizer( "AdamW" ) ) );

Pitch Sigma

Syntax: obj << Pitch Sigma( number=0 )

Description: Standard deviation of Gaussian pitch "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Pitch Sigma( 5 ) ) );

Pooling Layers

Syntax: obj << Pooling Layers( text=Max )

Description: Specifies pooling layers as a space-delimited list of one of four keywords: Max, Avg, Cat, or None. Last value carries forward if necessary. "Max" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Pooling Layers( "Max" ) ) );

Pretrained Tabular

Syntax: obj << Pretrained Tabular( ="None" )

Description: Specify a pretrained tabular model that is prepended to the Tabular Model. "None" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Pretrained Tabular( "None" ) ) );

Quantiles

Syntax: obj << Quantiles( text=0.9 )

Description: Specify a space-delimited list of quantiles to use for Quantile loss. "0.9" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Quantiles( "0.9" ) ) );

RPN NMS Threshold

Syntax: obj << RPN NMS Threshold( number=0.7 )

Description: Specifies the non-maximum suppression threshold for region proposals. Overlapping boxes with IOU values above this threshold are dropped. "0.7" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( RPN NMS Threshold( 0.7 ) ) );

Remove All But This Fit

Syntax: obj << ( fit[number] << Remove All But This Fit )

Description: Removes the reports and plots for all models except this one.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Iris.jmp" );
obj = dt << Torch Deep Learning(
    Y( :Species ),
    X( :Sepal length, :Sepal width, :Petal length, :Petal width ),
    Fit
);
Wait( 2 );
obj << (Fit[1] << Remove All But This Fit);

Remove Fit

Syntax: obj << ( fit[number] << Remove Fit )

Description: Removes the entire model report.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Iris.jmp" );
obj = dt << Torch Deep Learning(
    Y( :Species ),
    X( :Sepal length, :Sepal width, :Petal length, :Petal width ),
    Fit
);
Wait( 2 );
obj << (Fit[1] << Remove Fit);

Restore From

Syntax: obj << Restore From( " "=" " )

Description: Select a subfolder containing saved files from a previously fit model. Training for a new model will begin where this model finished. Model architectures and validation variables should match. Leave this field blank to train from scratch. " " by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Restore From( "" ) ) );

Roll Sigma

Syntax: obj << Roll Sigma( number=0 )

Description: Standard deviation of Gaussian roll "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Roll Sigma( 5 ) ) );

Save CAMs

Syntax: obj << Save CAMs

Description: Save gradient-based class activation maps (CAMs) as a new column.

JMP Version Added: 19

Save Embeddings

Syntax: obj << Save Embeddings

Description: Saves model embeddings (from final hidden layer) as new columns in the data table

JMP Version Added: 19

Save Model

Syntax: obj << Save Model

Description: Saves serialized modeling components to disk in a folder that you name. You can then specify this folder in Restore From to begin training with this model.

JMP Version Added: 19

Save Predicteds

Syntax: obj << Save Predicteds

Description: Saves the predicted values in a new column in the data table.

JMP Version Added: 19

Screening Method

Syntax: obj << Screening Method( "ResponseScreening"|"BootstrapForest"="ResponseScreening" )

Description: Choose a method by which to screen Tabular Model predictors prior to fitting the model within each fold. ResponseScreening is fast and BootstrapForest is more thorough. "ResponseScreening" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning(
    Y( :sex ),
    X( :picture ),
    Fit( Screening Method( "ResponseScreening" ) )
);

Screening Threshold

Syntax: obj << Screening Threshold( number=0 )

Description: If >= 1, the number of Tabular Model predictors to select by screening. If < 1, the predictors with cumulative portion less than the threshold. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Screening Threshold( 1 ) ) );

Seed

Syntax: obj << Seed( number=0 )

Description: Specifies the seed for the random number generator. Note results may not be fully reproducible with the same seed due to the stochastic nature of certain Torch calculations. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Seed( 0 ) ) );

Segmentation Model

Syntax: obj << Segmentation Model( "UNet"|"FPN"|"LinkNet"|"DeepLabV3"|"DeepLabV3Plus"|"PAN"|"PSPNet"="UNet" )

Description: Specifies the image segmentation model. "UNet" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/segmentation.jmp" );
Torch Deep Learning( Y( :Mask ), X( :Picture ), Sett( Segmentation Model( "VGG11_BN" ) ) );

Spline Order

Syntax: obj << Spline Order( number=3 )

Description: For Kolmogorov Arnold B Spline networks, specifies the order of the spline used for interpolation. "3" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Spline Order( 3 ) ) );

Strides

Syntax: obj << Strides( text=1 )

Description: For custom convolutional models, specifies the strides as a space-delimited list of positive integers. Last value carries forward if necessary. "1" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :height, :weight ), Fit( Strides( "1" ) ) );

Tabular Model

Syntax: obj << Tabular Model( "MultiLayerPerceptron"|"FTTransformer"|"KolmogorovArnoldBSpline"|"CustomConv1d"|"RadialBasisMachine"|"MixedModel"="MultiLayerPerceptron" )

Description: Specifies the tabular network architecture to use. Choose from Multilayer Perceptron (MLP), Feature Tokenized Transformer (FTTransformer), Kolmogorov Arnold Network (KolmogorovArnoldBSpline), or other options "MultiLayerPerceptron" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning(
    Y( :sex ),
    X( :picture ),
    Fit( Tabular Model( "MultiLayerPerceptron" ) )
);

Text Model

Syntax: obj << Text Model( ="BertTiny" )

Description: Specifies the text network architecture to use. Models are ordered by size. Smaller models train faster but may not perform as well as larger models. "BertTiny" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Chips.jmp" );
Torch Deep Learning(
    Y( :Buy again? ),
    X( :Potato Chip Product Review ),
    Fit( Text Model( "BERT" ) )
);

Triplet Loss Weight

Syntax: obj << Triplet Loss Weight( number=0.0 )

Description: Specifies the multiplier alpha to use in the following compound loss function: alpha * triplet_loss + (1 - alpha) * loss_function. Must be between 0 and 1. "0.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Triplet Loss Weight( 0.5 ) ) );

Use Data As Knots

Syntax: obj << Use Data As Knots( state=0 )

Description: For Radial Basis Machine models, check to use the training data as knots to form an interpolation-style model. "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning(
    Y( :sex, :height ),
    X( :picture ),
    Fit( Tabular Model( "Radial Basis Machine" ), Use Data As Knots( 1 ) )
);

VFlip Prob

Syntax: obj << VFlip Prob( number=0 )

Description: Probability of vertical flip "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( VFlip Prob( 0.2 ) ) );

Weight Decay

Syntax: obj << Weight Decay( number=0.0 )

Description: Specifies a penalty term multiplier of the L2 norm of the trainable parameters, which regularizes them in a way similar to ridge regression. "0.0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Weight Decay( 0.0001 ) ) );

Worker Count

Syntax: obj << Worker Count( number=4 )

Description: Specifies the number of workers to use to load batches of data during training. A number near half the number of actual cores is usually near optimal. "4" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Worker Count( 0 ) ) );

X Slide Sigma

Syntax: obj << X Slide Sigma( number=0 )

Description: Standard deviation of Gaussian random shift along the X axis "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( X Slide Sigma( 5 ) ) );

Y Slide Sigma

Syntax: obj << Y Slide Sigma( number=0 )

Description: Standard deviation of Gaussian random shift along the Y axis "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Y Slide Sigma( 5 ) ) );

Yaw Sigma

Syntax: obj << Yaw Sigma( number=0 )

Description: Standard deviation of Gaussian yaw "0" by default.

JMP Version Added: 19

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );
Torch Deep Learning( Y( :sex ), X( :picture ), Fit( Yaw Sigma( 5 ) ) );