search

SeldonIO / alibi

Algorithms for explaining machine learning models
https://docs.seldon.io/projects/alibi/en/stable/
Other
2.39k stars 249 forks source link

Custom input layers support for Integrated Gradients #287

Open gipster opened 4 years ago

gipster commented 4 years ago

At the moment the IntegratedGradients class only supports tf.keras and keras models taking numpy arrays as inputs and in which the input layer is a standard keras input layer such as tf.keras.Input or tf.keras.layers.InputLayer.

Investigation is ongoing to see whether it is possible to extend the method to models with custom input layers taking other types of inputs such as dictionaries.

Related issue

ofiryaish commented 4 years ago

Hey, what about multiple inputs to the model? I saw implantation for this but it doesn't support TF2. https://github.com/hiranumn/IntegratedGradients