DeepExplain | unified framework of perturbation | Machine Learning library

I managed to fix it by adding at the beginning of my code

The owner of the shap repo said:

The fundamental issue here is that DeepExplainer does not yet support TF 2.0.

That was on 11 Dec 2019. Is this still the case? Try it with Tensorflow 1.15 and see if that works.

Another issue on the shap repo about this (2 Jun 2020) says:

Alright, thank you. I did not see the post by Lundberg. I will stick to the workaround of using TF 1.15 until a new version of SHAP is released.

I think SHAP (whatever it is) is expecting a Numpy array and so indexing x_train like a Numpy array, it yields an error. Try:

The returned value of model.fit is not the model instance; rather, it's the history of training (i.e. stats like loss and metric values) as an instance of keras.callbacks.History class. That's why you get the mentioned error when you pass the returned History object to shap.DeepExplainer. Instead, you should pass the model instance itself:

Okay, via reading shap's source code, I realised it didn't take into account that the data were pandas' dataframes, eventhought the documentation says otherwise.

It worked using numpy.arrays

From the documentation, the first two arguments of shap.summary_plot are:

shap_values: numpy.array For single output explanations this is a matrix of SHAP values (# samples x # features). For multi-output explanations this is a list of such matrices of SHAP values.

features: numpy.array or pandas.DataFrame or list Matrix of feature values (# samples x # features) or a feature_names list as shorthand.

In the code that you provided, both shap_vals[0] and val_x[:10] have shape (10, 5, 4). Therefore, you should either flatten the first and second dimensions or select the timepoint you are interested in. For example: