Piecewise Linear Interpolator

Description

The piecewise linear interpolator fills the gaps between known data points using straight line segments, offering a straightforward way to connect the dots and obtain a complete picture of the data's progression.

Application

Piecewise linear interpolation is a method used often in calibration processes. Therefore, you could bring your measurement data and build an interpolator on a certain number of defined calibration points. The interpolator then can be used to test the quality of your calibration on other validation points.

How to use

In order to use this step, you will need a data set with at least 2 columns.

• In the Data field, select the data set that contains the data that will be used to create the interpolator.
• In the Input Column field, select the column which contains the input points that will be used. This column need to be sorted by ascending values (lowest to highest) prior to this step for the results to be meaningful. If this is not the case, you can use the step Sort by Columns.
• In the Output Column field, select the column which contains the output points that will be used.
• You can define the name of your model by editing the field Model Name.

To ensure that the interpolator is set up properly, please follow these additional instructions:

• The two selected columns need to be different,
• The values in the input column cannot be repeated (see Remove Duplicates),
• Ensure that the table doesn’t contain missing data (see Plot Missing Data, Remove Missing).

Once the piecewise linear interpolator is created, it can be used as a normal model in the platform (e.g. Scalar prediction, Dataset prediction, Targeted Optimisation, …)

Examples

In this example, 7 points were selected from a sine function, and a piecewise linear interpolator was created using these points. Once created, the model was evaluated on more points. The figure below shows a validation plot with the piecewise linear prediction in red and the true value in white.

More on this step

Currently, if the model is asked to make a prediction for an input that is out of the range initially used, it will return a constant value equal to the closest (lowest or highest) known input value.