Added shap

src/main.py

@@ -9,6 +9,7 @@ os.environ["TF_CPP_MIN_LOG_LEVEL"] = "1"
 import tensorflow as tf
 import keras_tuner as kt
 from keras import backend as K
+import shap 
 
 import numpy as np
 import matplotlib.pyplot as plt
@@ -154,7 +155,7 @@ def plot_importance(test_dataset):
     label_dataset = test_dataset.map(lambda x, y: y)
     test_dataset = test_dataset.map(lambda x, y: x)
     
-    el_num = (np.random.rand(4) * len_db * 0.075 * len(knots)).astype(int)
+    el_num = (np.random.rand(50) * len_db * 0.075 * len(knots)).astype(int)
 
     input_sequences = []
     input_labels = []
@@ -163,7 +164,17 @@ def plot_importance(test_dataset):
             input_sequences.append(x_element.numpy().reshape(1,-1))
             input_labels.append(y_element)
 
+    input_sequences = np.array(input_sequences)
+
+    explainer = shap.DeepExplainer(model, input_sequences.transpose(0,2,1))
+
     for j, (input_sequence, input_label) in enumerate(zip(input_sequences, input_labels)):
+        
+        plt.plot(explainer.shap_values(input_sequence.transpose(1,0)[None,:,:])[input_label].flatten())
+
+        plt.savefig(os.path.join(checkpoint_filepath,"layer_analysis",f"activation_gradient_label_{input_label}_test_{el_num[j]}.pdf"))
+        plt.cla()
+
         for i in range(len(model.layers)):
             intermediate_model=tf.keras.models.Model(inputs=model.input,outputs=model.layers[i].output) #Intermediate model between Input Layer and Output Layer which we are concerned about