Sensitivity analysis for a whole dataset

The sensitivity for a whole dataset is calculated as a norm of sensitivities on each sample in this dataset.

• Switch to Metrics panel

• Choose the file in Evaluation files section:

  • If the file was supplied earlier, it is already listed in Evaluation files

  • To add a new file for evaluation, press + in Additional section of Evaluation files

  • For Sensitivity analysis, the output file is not required

• Click on one of the output neurons (representing output features) on the plot of the neural network in the Network sensitivity section

• Each neuron becomes colored according to the sensitivity of the chosen output neurons with respect to this neuron

An example of the sensitivity analysis for a single sample:

Tabular network sensitivity for a whole dataset. Regression test case: Energy consumption.

Note

By default, the Network sensitivity is performed with the last model available in the checkpoint. Use the checkpoint slider in the bottom to choose any other available model.