Offline Trainer block

Description

The Offline Trainer block is used to train model blocks that are data driven, such as the Nonlinear Model block.

A database needs to selected for training the model, as this database is used as a training dataset. The Offline Trainer simply obtains a training dataset by extracting the input and target data from any data source to which it is connected, using only of good quality data. The training dataset uses input data and target data spanning the same time period. The dataset selected can be that data that you want modeled, or a different dataset altogether.   

Training of the model happens during the configuring of the Trainer block, when once the required fields have been configured, the [Train] function is selected. This takes a short while, depending on the size of your dataset selected.

Once the model has been trained, it remains effective until new outputs are added, or outputs are changed, at which time the model needs to be retrained.

This block is only used in design-time while the model is being trained; the block will not execute during run-time.

Return to Overview of blocks

the offline trainer block  

Block Type

Rules and Models block

Input port

There is a single input port to the Offline Trainer block, which MUST contain the target field(s).  The name of the target field(s) selected will have previously been configured during the configuring of the Model, and is merely selected from a populated list. The Offline Trainer will automatically check the blueprint data source, if configured, to determine whether the target field(s) are present.  If the target data is present in the blueprint data source, the Offline Trainer will be runnable, and clicking on [Train] will start the training process.      

Output port

There is no output port. This is because the Offline Trainer block performs a function on other blocks, and does not calculate any output values of its own.

Functions performed on tags

The offline trainer does not modify the values, timestamps or qualities of the target or input data in any way. It simply obtains a training dataset from the data source (composed only of good quality data) and runs the data through the model block.    

Return to top


Related topics:

  

CSense 2023- Last updated: June 24,2025