Build NeurEco Parametric Frequency Sweep model with the GUI

• Fill in the Settings tab, the build parameters are explained in the table below:

  

  Settings to build a Parametric Frequency Sweep model

• Press Build button

• Once the Build started, the Training, Evaluation, Metrics and Export panels become available. The moment the first model is saved to the checkpoint, these panels can be used as usual.

Build parameters

Minimum Settings to build a Parametric Frequency Sweep model

Name	Description
Training Data	Required. Data used to train a model. Click on Add files and choose paths to the files prepared according to Data preparation for NeurEco Parametric Frequency Sweep with the GUI
Validation Data	Optional. Data used to validate a model. If not provided, the Validation Data are chosen automatically among the provided samples in Training Data
Testing Data	Optional. Data never used during the training process. If provided, allow to monitor the model performance on the test data during the Build.
Minimum enrichment steps	Default = \(10\). Minimum number of enrichment steps to perform during the model construction.
Maximum enrichment steps	Default = \(200\). Maximum number of enrichment steps to perform during the model construction.
Unsuccessful enrichment steps	Default = \(4\). Stagnation criterium: tolerated number of subsequent enrichments without model improvement.
Validation percentage	Default = \(30.0\). Percentage of the data that NeurEco selects to use as Validation Data. Ignored when Validation Data are provided explicitly.

Advanced parameters

Advanced Settings to build a Parametric Frequency Sweep model

Name	Description
Reduced space dimension	Default = \(5\), Dimension of reduced space of outputs to use to train the model.
Enrichment rate	Float in range \([0,1]\), default = \(0.2\). Rate of enrichment. If is set to \(0\), the enrichment is conducted by one transformation at a time, il set to \(1\), all current possible transformations are performed together.

Data normalization for Parametric Frequency Sweep

NeurEco performs the data normalization automatically for Parametric Frequency Sweep.

• for input features: a Min-Max normalization is performed by feature, meaning that each input feature \(f\) is normalized independently from others, so that

  \[f_{normalized}=\frac{f-min(f)}{max(f)}\]

• for output features: all features are normalized together by division by their maximum absolute value, so that

  \[targets_{normalized}=\frac{targets}{max(|targets|)}\]