Lagun

Number of Inputs

Import Inputs

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Manually Change Inputs
(please save)

Warning: Updating inputs resets DOE.

Current Inputs

Export Inputs

Here you can change the outputs types (interest, control, status, constant).

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Define Objective Function

Open Panel:

Close Panel:

How optimizer is run

Select Optimizer

Export Problem Formulation

Current Inputs

Export Inputs

Select File

Browse...

Choose Normal/Expert Mode

Switching to expert mode enables more complex surrogate combinations to handle difficult problems.

It is highly recommended to try first this linear model. For some outputs it may be predictive enough and for the others you can check how nonlinear the output may be.

Choose Q2 LOO threshold

Only the outputs with the best surrogate so far less than the Q2 LOO threshold will be considered.

Choose Q2 LOO threshold

Only the outputs with the best surrogate so far less than the Q2 LOO threshold will be considered.

Interpolate ?

This option only applies to kriging surrogates. When the outputs are subjected to noise (i.e. experimental data), it is highly recommended to turn off the interpolation.

MultiStart Estim.

Choose Q2 LOO threshold

Only the outputs with the best surrogate so far less than the Q2 LOO threshold will be considered.

Choose Q2 LOO threshold

Only the outputs with the best surrogate so far less than the Q2 LOO threshold will be considered.

Click on a cell in the Q2 table below to display the prediction quality for each output and each metamodel.

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select Test File

Browse...

Here you can choose the criterion used to automatically assign the best surrogate model to each output.

If you make a manual change in the list of selected metamodels, it will not be saved unless you click on the save button below.

Improvement Criteria

Number of Additional Simulations

Refinement criteria not available for selected surrogate model

Import Inputs

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Manually Change Inputs
(please save)

If you change cutoffs, you need to reactive display to update the table

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select File

Browse...

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select File

Browse...

Define Inputs to Predict

Import File

Manual

Predict

Export Predictions

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Uncertainty Parameters

Import Inputs

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Manually Change Inputs
(please save)

If you do not import a file (or the imported file is not valid), by default each input will follow a uniform distribution on the domain given by the imported bounds (for continuous ones) or the levels (for categorical ones).
If you do not know the distribution of some inputs but want to estimate it with available observations, you can change their type to 'Estimated' and fit their distribution in the following panel.

You can specify groups of numeric inputs which are dependent. If you know the groups you can then either define all the group copulas or estimate them with a dataset. If you do not know the groups, you can also estimate them with available observations in the following panel.

Current Inputs

Export Inputs

Choose UQ propagation visualization

Sample Size

It is recommended to choose a sample size at least equal to 100/p if the probability to be estimated is of the order p. If you do not have an estimate of the probability, you can use the UQ propagation above as a rough estimate.

Resample ?

Refine Surrogate Model Near Threshold

Number of Additional Simulations

Export Additional Simulations

Refinement criteria not available for the selected surrogate model

Send the additional simulations to the "importDOE" panel where they can be manually launched.

Automatically launch the additional simulations with the current simulator settings.

Type

Import Inputs

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Manually Change Inputs
(please save)

For mono-objective optimization, the local BFGS algorithm is used.

For multi-objective optimization, the problem is solved with the genetic algorithm NSGAII.

Check Optimization Results

Number of Confirmation Points

Current Inputs

Export Inputs

Send the additional simulations to the "importDOE" panel where they can be manually launched.

Automatically launch the additional simulations with the current simulator settings.

Export Problem Formulation

Current Inputs

Export Inputs

Import Problem Formulation

Browse...

Import Inputs

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Manually Change Inputs
(please save)

Please select what to display & export if no optimum respecting all constraints was found.

Select Filter

Export Optimization Results

Check Optimization Results

Number of Confirmation Points

Send the additional simulations to the "importDOE" panel where they can be manually launched.

Automatically launch the additional simulations with the current simulator settings.

Export Problem Formulation

Current Inputs

Export Inputs

Import Problem Formulation

Browse...

Import Inputs

Separator

, (comma)

; (semi-colon)

Tab

Decimal

. (point)

, (comma)

Select file

Browse...

Manually Change Inputs
(please save)

Choose Max. Iterations

Choose Nb Pts / Iter

The next point to simulate is very close to the current DOE : the optimization may have converged.

Cannot find next points to simulate. If multiple points by iteration, this error may come from convergence problems of the metamodel update between iterations. Try to restart optimization with only one point by iteration.

Cannot update metamodel with the new simulated points : the optimization may have converged.

Import Problem Formulation

Browse...

Estimation of Problem Complexity

(% of points in the domain which satisfy the constraints with a given probability)

Select Robust Criterion

Constraint Probability

Export Optimization Results

Choose elements to save

Save or download?

Save in app

Download

Enter Filename

Download

DOE & Preliminary Exploration

Number of MC DOE for Comparison

Number of Resamples for ML Tool

Surrogate Models

Select Kriging Estimation Method

Number of Points Tested for Kriging Improvement

Grid Search Size for Advanced Models

GSA

Sample Size for Sobol

Sample Size for Shapley

Number of Resamples for GSA

Exploration

Number of Trajectories in Parallel Coordinates

Number of Steps for Sliders

Number of Significant Digits for Sliders

Range Factor for Y-Axis (% of data range)

Number of Contour Levels

Loading Lagun...

LAGUN

LAGUN is a R/Shiny platform providing a user-friendly interface to methods and algorithms dedicated to the exploration and the analysis of datasets. Guided workflows are provided to help non-expert users to apply safely the proposed methodologies.

History

The platform is organized in tabs, each one of them corresponding to a step above.

Click to expand the panels below to learn more through tutorials and test cases (links to Gitlab).

General Layout & Color Code

Prepare DOE

Problem Definition

Surrogate Model

Explore

Optimize

More

Advanced Settings

About

Change Inputs

Import Inputs

Manually Change Inputs (please save)

Warning: Updating inputs resets DOE.

Evaluate DOE

Current Inputs

Mouse Mode

Palette Colors

Represention

Export

Configure Simulator

Change Output Groups

Import Output Group

Manually Change Outputs (please save)

Outputs of interest will be available for surrogate modelling and will always be displayed by default in visual explorations.

Control outputs will be available for surrogate modelling but will be hidden by default in visual explorations.

Status outputs only give information about the loaded observations and cannot be approximated with a surrogate.

Constant outputs are outputs with very small variations and will be handled with simple surrogate models automatically without user intervention. By default they will not be displayed in visual explorations.

Here you can change the outputs types (interest, control, status, constant).

Import Output Discretization

Configure Simulator

Use a previous optimization

To perform this step, some results must be available. We suggest that you run the initial point (see next section 'Define Optimization Problem').

Change Weights

Change Initial Inputs Values

Current Inputs

Define Problem Formulation

Manually Change Formulation (please save)

Change Inputs

Import Inputs

Manually Change Inputs (please save)

Change Weights

View Currently Used DOE

Functional plot

Regression plot - One by One

Parallel coordinate plot - Scatter Plot Matrix

Regression plot - All in One

Layout

Palette Colors

Parallel Coordinate Plot

Center Plot

Bounds

Scatter Plot Matrix

Cluster Observations

Correlation

Generalized Correlation

Rules

Download Results

View Currently Used DOE

Confirm DOE Refresh

Refreshing the DOE will delete all the metamodels trained so far.

Choose Normal/Expert Mode

Switching to expert mode enables more complex surrogate combinations to handle difficult problems.

It is highly recommended to try first this linear model. For some outputs it may be predictive enough and for the others you can check how nonlinear the output may be.

Acosso1 Settings

You can choose to train the surrogate only on the outputs for which the Q2 of a previous surrogate was not sufficient enough. When the number of outputs is large and previous surrogates are of good quality, this accelerates the search of the best surrogate.

Only the outputs with the best surrogate so far less than the Q2 LOO threshold will be considered.

Acosso2 Settings

You can choose to train the surrogate only on the outputs for which the Q2 of a previous surrogate was not sufficient enough. When the number of outputs is large and previous surrogates are of good quality, this accelerates the search of the best surrogate.

Only the outputs with the best surrogate so far less than the Q2 LOO threshold will be considered.

Kriging Settings

This option only applies to kriging surrogates. When the outputs are subjected to noise (i.e. experimental data), it is highly recommended to turn off the interpolation.

This model is the most flexible one and can model complex input-outputs relationships. But if the number of inputs is larger than 15-20, its flexibility implies that its calibration may be hard and result in a poor surrogate.

For such situations, when activating the expert mode, it is possible to break down the calibration by using a trend given by one of the previously trained surrogate, and try to estimate the residuals only. To go further, one can also limit the variables involved in the kriging model.

You can choose to train the surrogate only on the outputs for which the Q2 of a previous surrogate was not sufficient enough. When the number of outputs is large and previous surrogates are of good quality, this accelerates the search of the best surrogate.

LAGUN is a R/Shiny platform providing a user-friendly interface to methods and algorithms dedicated to the exploration and the analysis of datasets.
Guided workflows are provided to help non-expert users to apply safely the proposed methodologies.

Manually Change Inputs
(please save)

Manually Change Outputs
(please save)

Manually Change Formulation
(please save)

Manually Change Inputs
(please save)

Manually Change Inputs
(please save)