shapley values - error on dimensions of the data

Question

I get the following error when running my code using shap for my CNN model: AssertionError: AssertionError:Instance must have 1 or 2 dimensions!` - I would appreciate any advice

Using 1484 background data samples could cause slower run times. Consider using shap.sample(data, K) or shap.kmeans(data, K) to summarize the background as K samples.

AssertionError Traceback (most recent call last) /tmp/ipykernel_193211/2006523994.py in <cell line: 7>() 5 # Generate SHAP values for a random test image 6 # shape (sample, height, width, channels) ----> 7 shap_values = explainer.shap_values(X_test) 8 shap.initjs() 9 # Plot the SHAP values for the test image

~/anaconda3/envs/tensorflow2_p310/lib/python3.10/site-packages/shap/explainers/_kernel.py in shap_values(self, X, **kwargs) 154 X = X.tolil() 155 assert x_type.endswith(arr_type) or sp.sparse.isspmatrix_lil(X), "Unknown instance type: " + x_type --> 156 assert len(X.shape) == 1 or len(X.shape) == 2, "Instance must have 1 or 2 dimensions!" 157 158 # single instance * AssertionError: Instance must have 1 or 2 dimensions!

My code: import shap # Define the SHAP explainer object explainer = shap.KernelExplainer(best_model, X_test)

# Generate SHAP values for a test image
# shape (nrsamples, height, width, channels)
shap_values = explainer.shap_values(X_test)
shap.initjs()
# Plot the SHAP values for the test image
shap.summary_plot(shap_values, X_test)

Dimensions of X_test: X_test.shape (1484, 74, 62, 119)

Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. — Community, May 19 '23 at 23:52

shapley values - error on dimensions of the data

Using 1484 background data samples could cause slower run times. Consider using shap.sample(data, K) or shap.kmeans(data, K) to summarize the background as K samples.

0 Answers0