Conversation
i did changess in feature importance and genrate visuaalization ,also add tensorflow
download scikeras library for using Keras Classifier and keras Regressor
SHAP Calculation for TensorFlow Models: Modified calculate_shap_values to use shap.DeepExplainer for TensorFlow models. Handling SHAP Output: Ensured that the output from shap.DeepExplainer is correctly wrapped in a shap.Explanation object for consistency.
Handling Predictions from TensorFlow Models: TensorFlow models return predictions differently. For regression, they output continuous values. For classification, they might output probabilities or logits. We adjust y_pred accordingly, converting probabilities to class labels when necessary. Model Type Parameter: Added model_type parameter to indicate whether the model is a TensorFlow or scikit-learn model. Default is 'sklearn' for backward compatibility. Classification Handling: For binary classification, we threshold probabilities at 0.5. For multi-class classification, we use np.argmax to get class labels.
please install scikeras libarary
There was a problem hiding this comment.
request change no 1:
in utils, there are some functions are missing, that's why CI was failed.
request change no 2:
also create a new file same as main.py, but this time, use tensorflow model imports instead of scikitlearn models we used earlier
in following snippets:
here scikit learn models has been used, same you have to use tensorflow models
here's the reference for tensorflow models: https://github.com/tensorflow/models/tree/master/official
|
example tf model (neural network models): You may tryout with smaller models. Note: have the previous scikitlearn support as it is, just in core.py add tensorflow models support, like when user use any tensorflow models in usage it'll be executable |
2 participants
No description provided.