How to use the package with one dimensional data?

[hmars-t]

Is it possible to apply the method of activation maximization also on a 1D-CNN with one dimensional data?

[keisen]

Yes, of course! However, you have to adjust some parameters for one dimensional data, because, in default, the parameters of ActivationMaximization is tuned for 2D-CNN model.

1. Modify input_range, input_modifiers and regularizers for your data properly.
2. Tune parameters (optimizer, input_modifiers, regularizers, or so on) to generagte the model inputs that maximize the output of the given score functions.

Please see below for details:

• tf_keras_vis.activation_maximization.ActivationMaximization.call()
• example for visualize_dense_layer

Please let's me know if you face any error.
Thanks!