Genetic Algorithm PLS Learner

Uses a genetic algorithm to optimize a PLS model for the input database. The best model will be written to the model port. A ranked list is written to the second out port. Use the Row To Model node to convert the results into KNIME models.

Options

Activity Field
The database field containing the data to be fitted
Descriptors
Select all numerical descriptor fields that should be included in the model generation
3D Grid Fields
Select all 3D grid fields that should be included in the model generation
Variance Threshold
Low variance filter threshold
Component Limit
At most the specified number of largest principal components will be used.
Condition Limit
Maximum PCA transform matrix condition number
Smooth
This field controls the width of the smoothing distribution used in accumulating the distributions during clustering. It is the standard deviation of a normal distribution
Population Size
Number of individuals in each generation. Each individual is a vector of descriptors that are used to build the model.
Genes per Individual
Number of descriptors per individual
Failure Limit
When a new generation is created the number of tries to create a new individual before the algorithm gives up and proceeds. Might lead to populations smaller than the given limit.
Point Mutations (%)
Percent of each generation consisting of point mutations of the previous generation members
Reproduction (%)
Top percent of each generation that is transferred into the next generation without change
Figure Of Merit
Figure of merit. R2 is the regression coefficient. Q2 is the cross-validated regression coefficient (Leave one out). The third option is meant for optimizing the regression coefficient and minizing the difference between regression coefficient and cross-validated regression coefficient simultaneously. Ideally R2 and Q2 should be identical. A gap between them is an indicator for overfitting. R2 and Q2 are also written to the output table.
Generation Linit
Maximum number of generation until the algorithm terminates.
Termination Criterion
Number of generation without change in the figure of merit before the algorithm terminates.
Number of Models
Number of models to export

Input Ports

Icon
Molecules with Activity Data

Output Ports

Icon
Best Model
Icon
Models

Popular Predecessors

  • No recommendations found

Popular Successors

  • No recommendations found

Views

Genetic Algorithm PLS Learner
Summary

Workflows

  • No workflows found

Links

Developers

You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.