diff --git a/choice_learn/models/base_model.py b/choice_learn/models/base_model.py
index ec6ecd44..d95c0bf5 100644
--- a/choice_learn/models/base_model.py
+++ b/choice_learn/models/base_model.py
@@ -444,7 +444,7 @@ def fit(
         self.callbacks.on_train_end(logs=temps_logs)
         return losses_history
 
-    @tf.function
+    @tf.function(reduce_retracing=True)
     def batch_predict(
         self,
         shared_features_by_choice,
@@ -731,7 +731,6 @@ def f(params_1d):
             # calculate gradients and convert to 1D tf.Tensor
             grads = tape.gradient(loss_value, self.trainable_weights)
             grads = tf.dynamic_stitch(idx, grads)
-            # print out iteration & loss
             f.iter.assign_add(1)
 
             # store loss value so we can retrieve later
diff --git a/choice_learn/models/latent_class_base_model.py b/choice_learn/models/latent_class_base_model.py
index 2f9321af..0c11645d 100644
--- a/choice_learn/models/latent_class_base_model.py
+++ b/choice_learn/models/latent_class_base_model.py
@@ -104,12 +104,18 @@ def instantiate(self, **kwargs):
             name="Latent-Logits",
         )
         self.latent_logits = init_logit
-        self.models = [self.model_class(**mp) for mp in self.model_parameters]
-        for model in self.models:
-            model.instantiate(**kwargs)
 
+        self.models = self.instantiate_latent_models(**kwargs)
         self.instantiated = True
 
+    def instantiate_latent_models(self, **kwargs):
+        """Instantiate latent models."""
+        models = [self.model_class(**mp) for mp in self.model_parameters]
+        for model in models:
+            model.instantiate(**kwargs)
+
+        return models
+
     # @tf.function
     def batch_predict(
         self,
@@ -249,7 +255,6 @@ def fit(self, choice_dataset, sample_weight=None, verbose=0):
         """
         if self.fit_method.lower() == "em":
             self.minf = np.log(1e-3)
-            print("Expectation-Maximization estimation algorithm not well implemented yet.")
             return self._em_fit(
                 choice_dataset=choice_dataset, sample_weight=sample_weight, verbose=verbose
             )
@@ -824,7 +829,7 @@ def _expectation(self, choice_dataset):
         )
 
         return tf.clip_by_value(
-            predicted_probas / np.sum(predicted_probas, axis=1, keepdims=True), 1e-10, 1
+            predicted_probas / np.sum(predicted_probas, axis=1, keepdims=True), 1e-6, 1
         ), loss
 
     def _maximization(self, choice_dataset, verbose=0):
@@ -842,10 +847,17 @@ def _maximization(self, choice_dataset, verbose=0):
         np.ndarray
             latent probabilities resulting of maximization step
         """
-        self.models = [self.model_class(**mp) for mp in self.model_parameters]
+        # models = [self.model_class(**mp) for mp in self.model_parameters]
+        # for i in range(len(models)):
+        #     for j, var in enumerate(self.models[i].trainable_weights):
+        #         models[i]._trainable_weights[j] = var
+        # self.instantiate_latent_models(choice_dataset)
+
         # M-step: MNL estimation
         for q in range(self.n_latent_classes):
-            self.models[q].fit(choice_dataset, sample_weight=self.weights[:, q], verbose=verbose)
+            self.models[q].fit(
+                choice_dataset, sample_weight=self.weights[:, q].numpy(), verbose=verbose
+            )
 
         # M-step: latent probability estimation
         latent_probas = np.sum(self.weights, axis=0)
@@ -876,7 +888,9 @@ def _em_fit(self, choice_dataset, sample_weight=None, verbose=0):
 
         # Initialization
         init_sample_weight = np.random.rand(self.n_latent_classes, len(choice_dataset))
-        init_sample_weight = init_sample_weight / np.sum(init_sample_weight, axis=0, keepdims=True)
+        init_sample_weight = np.clip(
+            init_sample_weight / np.sum(init_sample_weight, axis=0, keepdims=True), 1e-6, 1
+        )
         for i, model in enumerate(self.models):
             # model.instantiate()
             model.fit(choice_dataset, sample_weight=init_sample_weight[i], verbose=verbose)
@@ -888,7 +902,7 @@ def _em_fit(self, choice_dataset, sample_weight=None, verbose=0):
             if np.sum(np.isnan(self.latent_logits)) > 0:
                 print("Nan in logits")
                 break
-        return hist_logits, hist_loss
+        return hist_loss, hist_logits
 
     def predict_probas(self, choice_dataset, batch_size=-1):
         """Predicts the choice probabilities for each choice and each product of a ChoiceDataset.
diff --git a/choice_learn/models/latent_class_mnl.py b/choice_learn/models/latent_class_mnl.py
index 1c5b6b83..c53e0478 100644
--- a/choice_learn/models/latent_class_mnl.py
+++ b/choice_learn/models/latent_class_mnl.py
@@ -4,6 +4,8 @@
 
 import tensorflow as tf
 
+import choice_learn.tf_ops as tf_ops
+
 from .conditional_logit import ConditionalLogit, MNLCoefficients
 from .latent_class_base_model import BaseLatentClassModel
 from .simple_mnl import SimpleMNL
@@ -23,6 +25,7 @@ def __init__(
         intercept=None,
         optimizer="Adam",
         lr=0.001,
+        epochs_maximization=1000,
         **kwargs,
     ):
         """Initialize model.
@@ -56,7 +59,7 @@ def __init__(
             "batch_size": batch_size,
             "lbfgs_tolerance": lbfgs_tolerance,
             "lr": lr,
-            "epochs": 1000,
+            "epochs": epochs_maximization,
         }
 
         super().__init__(
@@ -88,6 +91,15 @@ def instantiate_latent_models(self, n_items, n_shared_features, n_items_features
             model.indexes, model.weights = model.instantiate(
                 n_items, n_shared_features, n_items_features
             )
+            model.exact_nll = tf_ops.CustomCategoricalCrossEntropy(
+                from_logits=False,
+                label_smoothing=0.0,
+                sparse=False,
+                axis=-1,
+                epsilon=1e-25,
+                name="exact_categorical_crossentropy",
+                reduction="sum_over_batch_size",
+            )
             model.instantiated = True
 
     def instantiate(self, n_items, n_shared_features, n_items_features):
diff --git a/notebooks/models/latent_class_model.ipynb b/notebooks/models/latent_class_model.ipynb
index 82b4ed8f..70fd8c7b 100644
--- a/notebooks/models/latent_class_model.ipynb
+++ b/notebooks/models/latent_class_model.ipynb
@@ -120,14 +120,30 @@
      "text": [
       "Using L-BFGS optimizer, setting up .fit() function\n",
       "Using L-BFGS optimizer, setting up .fit() function\n",
-      "Using L-BFGS optimizer, setting up .fit() function\n"
+      "Using L-BFGS optimizer, setting up .fit() function\n",
+      "WARNING:tensorflow:5 out of the last 5 calls to <function pfor.<locals>.f at 0x7fa18fb41af0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
      ]
     },
     {
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "/var/folders/zz/r1py7zhj35q75v09h8_42nzh0000gp/T/ipykernel_67121/1263996749.py:4: MatplotlibDeprecationWarning: The get_cmap function was deprecated in Matplotlib 3.7 and will be removed two minor releases later. Use ``matplotlib.colormaps[name]`` or ``matplotlib.colormaps.get_cmap(obj)`` instead.\n",
+      "WARNING:tensorflow:5 out of the last 5 calls to <function pfor.<locals>.f at 0x7fa18fb41af0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "WARNING:tensorflow:6 out of the last 6 calls to <function pfor.<locals>.f at 0x7fa18fb92280> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "WARNING:tensorflow:6 out of the last 6 calls to <function pfor.<locals>.f at 0x7fa18fb92280> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.\n",
+      "/var/folders/zz/r1py7zhj35q75v09h8_42nzh0000gp/T/ipykernel_27459/1263996749.py:4: MatplotlibDeprecationWarning: The get_cmap function was deprecated in Matplotlib 3.7 and will be removed two minor releases later. Use ``matplotlib.colormaps[name]`` or ``matplotlib.colormaps.get_cmap(obj)`` instead.\n",
       "  cmap = mpl.cm.get_cmap(\"Set1\")\n"
      ]
     },
@@ -135,196 +151,196 @@
      "data": {
       "text/html": [
        "<style type=\"text/css\">\n",
-       "#T_279d5_row0_col0, #T_279d5_row0_col1, #T_279d5_row0_col2, #T_279d5_row0_col3, #T_279d5_row0_col4, #T_279d5_row0_col5, #T_279d5_row1_col0, #T_279d5_row1_col1, #T_279d5_row1_col2, #T_279d5_row1_col3, #T_279d5_row1_col4, #T_279d5_row1_col5, #T_279d5_row2_col0, #T_279d5_row2_col1, #T_279d5_row2_col2, #T_279d5_row2_col3, #T_279d5_row2_col4, #T_279d5_row2_col5, #T_279d5_row3_col0, #T_279d5_row3_col1, #T_279d5_row3_col2, #T_279d5_row3_col3, #T_279d5_row3_col4, #T_279d5_row3_col5, #T_279d5_row4_col0, #T_279d5_row4_col1, #T_279d5_row4_col2, #T_279d5_row4_col3, #T_279d5_row4_col4, #T_279d5_row4_col5, #T_279d5_row5_col0, #T_279d5_row5_col1, #T_279d5_row5_col2, #T_279d5_row5_col3, #T_279d5_row5_col4, #T_279d5_row5_col5 {\n",
+       "#T_ea03d_row0_col0, #T_ea03d_row0_col1, #T_ea03d_row0_col2, #T_ea03d_row0_col3, #T_ea03d_row0_col4, #T_ea03d_row0_col5, #T_ea03d_row1_col0, #T_ea03d_row1_col1, #T_ea03d_row1_col2, #T_ea03d_row1_col3, #T_ea03d_row1_col4, #T_ea03d_row1_col5, #T_ea03d_row2_col0, #T_ea03d_row2_col1, #T_ea03d_row2_col2, #T_ea03d_row2_col3, #T_ea03d_row2_col4, #T_ea03d_row2_col5, #T_ea03d_row3_col0, #T_ea03d_row3_col1, #T_ea03d_row3_col2, #T_ea03d_row3_col3, #T_ea03d_row3_col4, #T_ea03d_row3_col5, #T_ea03d_row4_col0, #T_ea03d_row4_col1, #T_ea03d_row4_col2, #T_ea03d_row4_col3, #T_ea03d_row4_col4, #T_ea03d_row4_col5, #T_ea03d_row5_col0, #T_ea03d_row5_col1, #T_ea03d_row5_col2, #T_ea03d_row5_col3, #T_ea03d_row5_col4, #T_ea03d_row5_col5 {\n",
        "  background-color: #e41a1c;\n",
        "}\n",
-       "#T_279d5_row6_col0, #T_279d5_row6_col1, #T_279d5_row6_col2, #T_279d5_row6_col3, #T_279d5_row6_col4, #T_279d5_row6_col5, #T_279d5_row7_col0, #T_279d5_row7_col1, #T_279d5_row7_col2, #T_279d5_row7_col3, #T_279d5_row7_col4, #T_279d5_row7_col5, #T_279d5_row8_col0, #T_279d5_row8_col1, #T_279d5_row8_col2, #T_279d5_row8_col3, #T_279d5_row8_col4, #T_279d5_row8_col5, #T_279d5_row9_col0, #T_279d5_row9_col1, #T_279d5_row9_col2, #T_279d5_row9_col3, #T_279d5_row9_col4, #T_279d5_row9_col5, #T_279d5_row10_col0, #T_279d5_row10_col1, #T_279d5_row10_col2, #T_279d5_row10_col3, #T_279d5_row10_col4, #T_279d5_row10_col5, #T_279d5_row11_col0, #T_279d5_row11_col1, #T_279d5_row11_col2, #T_279d5_row11_col3, #T_279d5_row11_col4, #T_279d5_row11_col5 {\n",
+       "#T_ea03d_row6_col0, #T_ea03d_row6_col1, #T_ea03d_row6_col2, #T_ea03d_row6_col3, #T_ea03d_row6_col4, #T_ea03d_row6_col5, #T_ea03d_row7_col0, #T_ea03d_row7_col1, #T_ea03d_row7_col2, #T_ea03d_row7_col3, #T_ea03d_row7_col4, #T_ea03d_row7_col5, #T_ea03d_row8_col0, #T_ea03d_row8_col1, #T_ea03d_row8_col2, #T_ea03d_row8_col3, #T_ea03d_row8_col4, #T_ea03d_row8_col5, #T_ea03d_row9_col0, #T_ea03d_row9_col1, #T_ea03d_row9_col2, #T_ea03d_row9_col3, #T_ea03d_row9_col4, #T_ea03d_row9_col5, #T_ea03d_row10_col0, #T_ea03d_row10_col1, #T_ea03d_row10_col2, #T_ea03d_row10_col3, #T_ea03d_row10_col4, #T_ea03d_row10_col5, #T_ea03d_row11_col0, #T_ea03d_row11_col1, #T_ea03d_row11_col2, #T_ea03d_row11_col3, #T_ea03d_row11_col4, #T_ea03d_row11_col5 {\n",
        "  background-color: #377eb8;\n",
        "}\n",
-       "#T_279d5_row12_col0, #T_279d5_row12_col1, #T_279d5_row12_col2, #T_279d5_row12_col3, #T_279d5_row12_col4, #T_279d5_row12_col5, #T_279d5_row13_col0, #T_279d5_row13_col1, #T_279d5_row13_col2, #T_279d5_row13_col3, #T_279d5_row13_col4, #T_279d5_row13_col5, #T_279d5_row14_col0, #T_279d5_row14_col1, #T_279d5_row14_col2, #T_279d5_row14_col3, #T_279d5_row14_col4, #T_279d5_row14_col5, #T_279d5_row15_col0, #T_279d5_row15_col1, #T_279d5_row15_col2, #T_279d5_row15_col3, #T_279d5_row15_col4, #T_279d5_row15_col5, #T_279d5_row16_col0, #T_279d5_row16_col1, #T_279d5_row16_col2, #T_279d5_row16_col3, #T_279d5_row16_col4, #T_279d5_row16_col5, #T_279d5_row17_col0, #T_279d5_row17_col1, #T_279d5_row17_col2, #T_279d5_row17_col3, #T_279d5_row17_col4, #T_279d5_row17_col5 {\n",
+       "#T_ea03d_row12_col0, #T_ea03d_row12_col1, #T_ea03d_row12_col2, #T_ea03d_row12_col3, #T_ea03d_row12_col4, #T_ea03d_row12_col5, #T_ea03d_row13_col0, #T_ea03d_row13_col1, #T_ea03d_row13_col2, #T_ea03d_row13_col3, #T_ea03d_row13_col4, #T_ea03d_row13_col5, #T_ea03d_row14_col0, #T_ea03d_row14_col1, #T_ea03d_row14_col2, #T_ea03d_row14_col3, #T_ea03d_row14_col4, #T_ea03d_row14_col5, #T_ea03d_row15_col0, #T_ea03d_row15_col1, #T_ea03d_row15_col2, #T_ea03d_row15_col3, #T_ea03d_row15_col4, #T_ea03d_row15_col5, #T_ea03d_row16_col0, #T_ea03d_row16_col1, #T_ea03d_row16_col2, #T_ea03d_row16_col3, #T_ea03d_row16_col4, #T_ea03d_row16_col5, #T_ea03d_row17_col0, #T_ea03d_row17_col1, #T_ea03d_row17_col2, #T_ea03d_row17_col3, #T_ea03d_row17_col4, #T_ea03d_row17_col5 {\n",
        "  background-color: #4daf4a;\n",
        "}\n",
        "</style>\n",
-       "<table id=\"T_279d5\">\n",
+       "<table id=\"T_ea03d\">\n",
        "  <thead>\n",
        "    <tr>\n",
        "      <th class=\"blank level0\" >&nbsp;</th>\n",
-       "      <th id=\"T_279d5_level0_col0\" class=\"col_heading level0 col0\" >Latent Class</th>\n",
-       "      <th id=\"T_279d5_level0_col1\" class=\"col_heading level0 col1\" >Coefficient Name</th>\n",
-       "      <th id=\"T_279d5_level0_col2\" class=\"col_heading level0 col2\" >Coefficient Estimation</th>\n",
-       "      <th id=\"T_279d5_level0_col3\" class=\"col_heading level0 col3\" >Std. Err</th>\n",
-       "      <th id=\"T_279d5_level0_col4\" class=\"col_heading level0 col4\" >z_value</th>\n",
-       "      <th id=\"T_279d5_level0_col5\" class=\"col_heading level0 col5\" >P(.>z)</th>\n",
+       "      <th id=\"T_ea03d_level0_col0\" class=\"col_heading level0 col0\" >Latent Class</th>\n",
+       "      <th id=\"T_ea03d_level0_col1\" class=\"col_heading level0 col1\" >Coefficient Name</th>\n",
+       "      <th id=\"T_ea03d_level0_col2\" class=\"col_heading level0 col2\" >Coefficient Estimation</th>\n",
+       "      <th id=\"T_ea03d_level0_col3\" class=\"col_heading level0 col3\" >Std. Err</th>\n",
+       "      <th id=\"T_ea03d_level0_col4\" class=\"col_heading level0 col4\" >z_value</th>\n",
+       "      <th id=\"T_ea03d_level0_col5\" class=\"col_heading level0 col5\" >P(.>z)</th>\n",
        "    </tr>\n",
        "  </thead>\n",
        "  <tbody>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row0\" class=\"row_heading level0 row0\" >0</th>\n",
-       "      <td id=\"T_279d5_row0_col0\" class=\"data row0 col0\" >0</td>\n",
-       "      <td id=\"T_279d5_row0_col1\" class=\"data row0 col1\" >Weights_items_features_0</td>\n",
-       "      <td id=\"T_279d5_row0_col2\" class=\"data row0 col2\" >-0.675645</td>\n",
-       "      <td id=\"T_279d5_row0_col3\" class=\"data row0 col3\" >0.023987</td>\n",
-       "      <td id=\"T_279d5_row0_col4\" class=\"data row0 col4\" >-28.167109</td>\n",
-       "      <td id=\"T_279d5_row0_col5\" class=\"data row0 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row0\" class=\"row_heading level0 row0\" >0</th>\n",
+       "      <td id=\"T_ea03d_row0_col0\" class=\"data row0 col0\" >0</td>\n",
+       "      <td id=\"T_ea03d_row0_col1\" class=\"data row0 col1\" >Weights_items_features_0</td>\n",
+       "      <td id=\"T_ea03d_row0_col2\" class=\"data row0 col2\" >-0.675645</td>\n",
+       "      <td id=\"T_ea03d_row0_col3\" class=\"data row0 col3\" >0.023987</td>\n",
+       "      <td id=\"T_ea03d_row0_col4\" class=\"data row0 col4\" >-28.167109</td>\n",
+       "      <td id=\"T_ea03d_row0_col5\" class=\"data row0 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row1\" class=\"row_heading level0 row1\" >1</th>\n",
-       "      <td id=\"T_279d5_row1_col0\" class=\"data row1 col0\" >0</td>\n",
-       "      <td id=\"T_279d5_row1_col1\" class=\"data row1 col1\" >Weights_items_features_1</td>\n",
-       "      <td id=\"T_279d5_row1_col2\" class=\"data row1 col2\" >-0.060604</td>\n",
-       "      <td id=\"T_279d5_row1_col3\" class=\"data row1 col3\" >0.008162</td>\n",
-       "      <td id=\"T_279d5_row1_col4\" class=\"data row1 col4\" >-7.424849</td>\n",
-       "      <td id=\"T_279d5_row1_col5\" class=\"data row1 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row1\" class=\"row_heading level0 row1\" >1</th>\n",
+       "      <td id=\"T_ea03d_row1_col0\" class=\"data row1 col0\" >0</td>\n",
+       "      <td id=\"T_ea03d_row1_col1\" class=\"data row1 col1\" >Weights_items_features_1</td>\n",
+       "      <td id=\"T_ea03d_row1_col2\" class=\"data row1 col2\" >-0.060604</td>\n",
+       "      <td id=\"T_ea03d_row1_col3\" class=\"data row1 col3\" >0.008162</td>\n",
+       "      <td id=\"T_ea03d_row1_col4\" class=\"data row1 col4\" >-7.424849</td>\n",
+       "      <td id=\"T_ea03d_row1_col5\" class=\"data row1 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row2\" class=\"row_heading level0 row2\" >2</th>\n",
-       "      <td id=\"T_279d5_row2_col0\" class=\"data row2 col0\" >0</td>\n",
-       "      <td id=\"T_279d5_row2_col1\" class=\"data row2 col1\" >Weights_items_features_2</td>\n",
-       "      <td id=\"T_279d5_row2_col2\" class=\"data row2 col2\" >1.851951</td>\n",
-       "      <td id=\"T_279d5_row2_col3\" class=\"data row2 col3\" >0.054914</td>\n",
-       "      <td id=\"T_279d5_row2_col4\" class=\"data row2 col4\" >33.724579</td>\n",
-       "      <td id=\"T_279d5_row2_col5\" class=\"data row2 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row2\" class=\"row_heading level0 row2\" >2</th>\n",
+       "      <td id=\"T_ea03d_row2_col0\" class=\"data row2 col0\" >0</td>\n",
+       "      <td id=\"T_ea03d_row2_col1\" class=\"data row2 col1\" >Weights_items_features_2</td>\n",
+       "      <td id=\"T_ea03d_row2_col2\" class=\"data row2 col2\" >1.851951</td>\n",
+       "      <td id=\"T_ea03d_row2_col3\" class=\"data row2 col3\" >0.054914</td>\n",
+       "      <td id=\"T_ea03d_row2_col4\" class=\"data row2 col4\" >33.724579</td>\n",
+       "      <td id=\"T_ea03d_row2_col5\" class=\"data row2 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row3\" class=\"row_heading level0 row3\" >3</th>\n",
-       "      <td id=\"T_279d5_row3_col0\" class=\"data row3 col0\" >0</td>\n",
-       "      <td id=\"T_279d5_row3_col1\" class=\"data row3 col1\" >Weights_items_features_3</td>\n",
-       "      <td id=\"T_279d5_row3_col2\" class=\"data row3 col2\" >1.322549</td>\n",
-       "      <td id=\"T_279d5_row3_col3\" class=\"data row3 col3\" >0.048159</td>\n",
-       "      <td id=\"T_279d5_row3_col4\" class=\"data row3 col4\" >27.462420</td>\n",
-       "      <td id=\"T_279d5_row3_col5\" class=\"data row3 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row3\" class=\"row_heading level0 row3\" >3</th>\n",
+       "      <td id=\"T_ea03d_row3_col0\" class=\"data row3 col0\" >0</td>\n",
+       "      <td id=\"T_ea03d_row3_col1\" class=\"data row3 col1\" >Weights_items_features_3</td>\n",
+       "      <td id=\"T_ea03d_row3_col2\" class=\"data row3 col2\" >1.322549</td>\n",
+       "      <td id=\"T_ea03d_row3_col3\" class=\"data row3 col3\" >0.048159</td>\n",
+       "      <td id=\"T_ea03d_row3_col4\" class=\"data row3 col4\" >27.462420</td>\n",
+       "      <td id=\"T_ea03d_row3_col5\" class=\"data row3 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row4\" class=\"row_heading level0 row4\" >4</th>\n",
-       "      <td id=\"T_279d5_row4_col0\" class=\"data row4 col0\" >0</td>\n",
-       "      <td id=\"T_279d5_row4_col1\" class=\"data row4 col1\" >Weights_items_features_4</td>\n",
-       "      <td id=\"T_279d5_row4_col2\" class=\"data row4 col2\" >-5.857089</td>\n",
-       "      <td id=\"T_279d5_row4_col3\" class=\"data row4 col3\" >0.191162</td>\n",
-       "      <td id=\"T_279d5_row4_col4\" class=\"data row4 col4\" >-30.639460</td>\n",
-       "      <td id=\"T_279d5_row4_col5\" class=\"data row4 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row4\" class=\"row_heading level0 row4\" >4</th>\n",
+       "      <td id=\"T_ea03d_row4_col0\" class=\"data row4 col0\" >0</td>\n",
+       "      <td id=\"T_ea03d_row4_col1\" class=\"data row4 col1\" >Weights_items_features_4</td>\n",
+       "      <td id=\"T_ea03d_row4_col2\" class=\"data row4 col2\" >-5.857089</td>\n",
+       "      <td id=\"T_ea03d_row4_col3\" class=\"data row4 col3\" >0.191162</td>\n",
+       "      <td id=\"T_ea03d_row4_col4\" class=\"data row4 col4\" >-30.639460</td>\n",
+       "      <td id=\"T_ea03d_row4_col5\" class=\"data row4 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row5\" class=\"row_heading level0 row5\" >5</th>\n",
-       "      <td id=\"T_279d5_row5_col0\" class=\"data row5 col0\" >0</td>\n",
-       "      <td id=\"T_279d5_row5_col1\" class=\"data row5 col1\" >Weights_items_features_5</td>\n",
-       "      <td id=\"T_279d5_row5_col2\" class=\"data row5 col2\" >-6.513206</td>\n",
-       "      <td id=\"T_279d5_row5_col3\" class=\"data row5 col3\" >0.195680</td>\n",
-       "      <td id=\"T_279d5_row5_col4\" class=\"data row5 col4\" >-33.285046</td>\n",
-       "      <td id=\"T_279d5_row5_col5\" class=\"data row5 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row5\" class=\"row_heading level0 row5\" >5</th>\n",
+       "      <td id=\"T_ea03d_row5_col0\" class=\"data row5 col0\" >0</td>\n",
+       "      <td id=\"T_ea03d_row5_col1\" class=\"data row5 col1\" >Weights_items_features_5</td>\n",
+       "      <td id=\"T_ea03d_row5_col2\" class=\"data row5 col2\" >-6.513206</td>\n",
+       "      <td id=\"T_ea03d_row5_col3\" class=\"data row5 col3\" >0.195680</td>\n",
+       "      <td id=\"T_ea03d_row5_col4\" class=\"data row5 col4\" >-33.285046</td>\n",
+       "      <td id=\"T_ea03d_row5_col5\" class=\"data row5 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row6\" class=\"row_heading level0 row6\" >6</th>\n",
-       "      <td id=\"T_279d5_row6_col0\" class=\"data row6 col0\" >1</td>\n",
-       "      <td id=\"T_279d5_row6_col1\" class=\"data row6 col1\" >Weights_items_features_0</td>\n",
-       "      <td id=\"T_279d5_row6_col2\" class=\"data row6 col2\" >-1.817566</td>\n",
-       "      <td id=\"T_279d5_row6_col3\" class=\"data row6 col3\" >0.077771</td>\n",
-       "      <td id=\"T_279d5_row6_col4\" class=\"data row6 col4\" >-23.370796</td>\n",
-       "      <td id=\"T_279d5_row6_col5\" class=\"data row6 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row6\" class=\"row_heading level0 row6\" >6</th>\n",
+       "      <td id=\"T_ea03d_row6_col0\" class=\"data row6 col0\" >1</td>\n",
+       "      <td id=\"T_ea03d_row6_col1\" class=\"data row6 col1\" >Weights_items_features_0</td>\n",
+       "      <td id=\"T_ea03d_row6_col2\" class=\"data row6 col2\" >-1.817566</td>\n",
+       "      <td id=\"T_ea03d_row6_col3\" class=\"data row6 col3\" >0.077771</td>\n",
+       "      <td id=\"T_ea03d_row6_col4\" class=\"data row6 col4\" >-23.370796</td>\n",
+       "      <td id=\"T_ea03d_row6_col5\" class=\"data row6 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row7\" class=\"row_heading level0 row7\" >7</th>\n",
-       "      <td id=\"T_279d5_row7_col0\" class=\"data row7 col0\" >1</td>\n",
-       "      <td id=\"T_279d5_row7_col1\" class=\"data row7 col1\" >Weights_items_features_1</td>\n",
-       "      <td id=\"T_279d5_row7_col2\" class=\"data row7 col2\" >-1.726365</td>\n",
-       "      <td id=\"T_279d5_row7_col3\" class=\"data row7 col3\" >0.058838</td>\n",
-       "      <td id=\"T_279d5_row7_col4\" class=\"data row7 col4\" >-29.340986</td>\n",
-       "      <td id=\"T_279d5_row7_col5\" class=\"data row7 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row7\" class=\"row_heading level0 row7\" >7</th>\n",
+       "      <td id=\"T_ea03d_row7_col0\" class=\"data row7 col0\" >1</td>\n",
+       "      <td id=\"T_ea03d_row7_col1\" class=\"data row7 col1\" >Weights_items_features_1</td>\n",
+       "      <td id=\"T_ea03d_row7_col2\" class=\"data row7 col2\" >-1.726365</td>\n",
+       "      <td id=\"T_ea03d_row7_col3\" class=\"data row7 col3\" >0.058838</td>\n",
+       "      <td id=\"T_ea03d_row7_col4\" class=\"data row7 col4\" >-29.340986</td>\n",
+       "      <td id=\"T_ea03d_row7_col5\" class=\"data row7 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row8\" class=\"row_heading level0 row8\" >8</th>\n",
-       "      <td id=\"T_279d5_row8_col0\" class=\"data row8 col0\" >1</td>\n",
-       "      <td id=\"T_279d5_row8_col1\" class=\"data row8 col1\" >Weights_items_features_2</td>\n",
-       "      <td id=\"T_279d5_row8_col2\" class=\"data row8 col2\" >3.696567</td>\n",
-       "      <td id=\"T_279d5_row8_col3\" class=\"data row8 col3\" >0.160258</td>\n",
-       "      <td id=\"T_279d5_row8_col4\" class=\"data row8 col4\" >23.066404</td>\n",
-       "      <td id=\"T_279d5_row8_col5\" class=\"data row8 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row8\" class=\"row_heading level0 row8\" >8</th>\n",
+       "      <td id=\"T_ea03d_row8_col0\" class=\"data row8 col0\" >1</td>\n",
+       "      <td id=\"T_ea03d_row8_col1\" class=\"data row8 col1\" >Weights_items_features_2</td>\n",
+       "      <td id=\"T_ea03d_row8_col2\" class=\"data row8 col2\" >3.696567</td>\n",
+       "      <td id=\"T_ea03d_row8_col3\" class=\"data row8 col3\" >0.160258</td>\n",
+       "      <td id=\"T_ea03d_row8_col4\" class=\"data row8 col4\" >23.066404</td>\n",
+       "      <td id=\"T_ea03d_row8_col5\" class=\"data row8 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row9\" class=\"row_heading level0 row9\" >9</th>\n",
-       "      <td id=\"T_279d5_row9_col0\" class=\"data row9 col0\" >1</td>\n",
-       "      <td id=\"T_279d5_row9_col1\" class=\"data row9 col1\" >Weights_items_features_3</td>\n",
-       "      <td id=\"T_279d5_row9_col2\" class=\"data row9 col2\" >4.111840</td>\n",
-       "      <td id=\"T_279d5_row9_col3\" class=\"data row9 col3\" >0.157179</td>\n",
-       "      <td id=\"T_279d5_row9_col4\" class=\"data row9 col4\" >26.160225</td>\n",
-       "      <td id=\"T_279d5_row9_col5\" class=\"data row9 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row9\" class=\"row_heading level0 row9\" >9</th>\n",
+       "      <td id=\"T_ea03d_row9_col0\" class=\"data row9 col0\" >1</td>\n",
+       "      <td id=\"T_ea03d_row9_col1\" class=\"data row9 col1\" >Weights_items_features_3</td>\n",
+       "      <td id=\"T_ea03d_row9_col2\" class=\"data row9 col2\" >4.111840</td>\n",
+       "      <td id=\"T_ea03d_row9_col3\" class=\"data row9 col3\" >0.157179</td>\n",
+       "      <td id=\"T_ea03d_row9_col4\" class=\"data row9 col4\" >26.160225</td>\n",
+       "      <td id=\"T_ea03d_row9_col5\" class=\"data row9 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row10\" class=\"row_heading level0 row10\" >10</th>\n",
-       "      <td id=\"T_279d5_row10_col0\" class=\"data row10 col0\" >1</td>\n",
-       "      <td id=\"T_279d5_row10_col1\" class=\"data row10 col1\" >Weights_items_features_4</td>\n",
-       "      <td id=\"T_279d5_row10_col2\" class=\"data row10 col2\" >-26.693516</td>\n",
-       "      <td id=\"T_279d5_row10_col3\" class=\"data row10 col3\" >3.274723</td>\n",
-       "      <td id=\"T_279d5_row10_col4\" class=\"data row10 col4\" >-8.151381</td>\n",
-       "      <td id=\"T_279d5_row10_col5\" class=\"data row10 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row10\" class=\"row_heading level0 row10\" >10</th>\n",
+       "      <td id=\"T_ea03d_row10_col0\" class=\"data row10 col0\" >1</td>\n",
+       "      <td id=\"T_ea03d_row10_col1\" class=\"data row10 col1\" >Weights_items_features_4</td>\n",
+       "      <td id=\"T_ea03d_row10_col2\" class=\"data row10 col2\" >-26.693516</td>\n",
+       "      <td id=\"T_ea03d_row10_col3\" class=\"data row10 col3\" >3.274723</td>\n",
+       "      <td id=\"T_ea03d_row10_col4\" class=\"data row10 col4\" >-8.151381</td>\n",
+       "      <td id=\"T_ea03d_row10_col5\" class=\"data row10 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row11\" class=\"row_heading level0 row11\" >11</th>\n",
-       "      <td id=\"T_279d5_row11_col0\" class=\"data row11 col0\" >1</td>\n",
-       "      <td id=\"T_279d5_row11_col1\" class=\"data row11 col1\" >Weights_items_features_5</td>\n",
-       "      <td id=\"T_279d5_row11_col2\" class=\"data row11 col2\" >-14.925840</td>\n",
-       "      <td id=\"T_279d5_row11_col3\" class=\"data row11 col3\" >0.634699</td>\n",
-       "      <td id=\"T_279d5_row11_col4\" class=\"data row11 col4\" >-23.516403</td>\n",
-       "      <td id=\"T_279d5_row11_col5\" class=\"data row11 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row11\" class=\"row_heading level0 row11\" >11</th>\n",
+       "      <td id=\"T_ea03d_row11_col0\" class=\"data row11 col0\" >1</td>\n",
+       "      <td id=\"T_ea03d_row11_col1\" class=\"data row11 col1\" >Weights_items_features_5</td>\n",
+       "      <td id=\"T_ea03d_row11_col2\" class=\"data row11 col2\" >-14.925840</td>\n",
+       "      <td id=\"T_ea03d_row11_col3\" class=\"data row11 col3\" >0.634699</td>\n",
+       "      <td id=\"T_ea03d_row11_col4\" class=\"data row11 col4\" >-23.516403</td>\n",
+       "      <td id=\"T_ea03d_row11_col5\" class=\"data row11 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row12\" class=\"row_heading level0 row12\" >12</th>\n",
-       "      <td id=\"T_279d5_row12_col0\" class=\"data row12 col0\" >2</td>\n",
-       "      <td id=\"T_279d5_row12_col1\" class=\"data row12 col1\" >Weights_items_features_0</td>\n",
-       "      <td id=\"T_279d5_row12_col2\" class=\"data row12 col2\" >-2.104791</td>\n",
-       "      <td id=\"T_279d5_row12_col3\" class=\"data row12 col3\" >0.104296</td>\n",
-       "      <td id=\"T_279d5_row12_col4\" class=\"data row12 col4\" >-20.181009</td>\n",
-       "      <td id=\"T_279d5_row12_col5\" class=\"data row12 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row12\" class=\"row_heading level0 row12\" >12</th>\n",
+       "      <td id=\"T_ea03d_row12_col0\" class=\"data row12 col0\" >2</td>\n",
+       "      <td id=\"T_ea03d_row12_col1\" class=\"data row12 col1\" >Weights_items_features_0</td>\n",
+       "      <td id=\"T_ea03d_row12_col2\" class=\"data row12 col2\" >-2.104791</td>\n",
+       "      <td id=\"T_ea03d_row12_col3\" class=\"data row12 col3\" >0.104296</td>\n",
+       "      <td id=\"T_ea03d_row12_col4\" class=\"data row12 col4\" >-20.181009</td>\n",
+       "      <td id=\"T_ea03d_row12_col5\" class=\"data row12 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row13\" class=\"row_heading level0 row13\" >13</th>\n",
-       "      <td id=\"T_279d5_row13_col0\" class=\"data row13 col0\" >2</td>\n",
-       "      <td id=\"T_279d5_row13_col1\" class=\"data row13 col1\" >Weights_items_features_1</td>\n",
-       "      <td id=\"T_279d5_row13_col2\" class=\"data row13 col2\" >-1.652622</td>\n",
-       "      <td id=\"T_279d5_row13_col3\" class=\"data row13 col3\" >0.073820</td>\n",
-       "      <td id=\"T_279d5_row13_col4\" class=\"data row13 col4\" >-22.387188</td>\n",
-       "      <td id=\"T_279d5_row13_col5\" class=\"data row13 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row13\" class=\"row_heading level0 row13\" >13</th>\n",
+       "      <td id=\"T_ea03d_row13_col0\" class=\"data row13 col0\" >2</td>\n",
+       "      <td id=\"T_ea03d_row13_col1\" class=\"data row13 col1\" >Weights_items_features_1</td>\n",
+       "      <td id=\"T_ea03d_row13_col2\" class=\"data row13 col2\" >-1.652622</td>\n",
+       "      <td id=\"T_ea03d_row13_col3\" class=\"data row13 col3\" >0.073820</td>\n",
+       "      <td id=\"T_ea03d_row13_col4\" class=\"data row13 col4\" >-22.387188</td>\n",
+       "      <td id=\"T_ea03d_row13_col5\" class=\"data row13 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row14\" class=\"row_heading level0 row14\" >14</th>\n",
-       "      <td id=\"T_279d5_row14_col0\" class=\"data row14 col0\" >2</td>\n",
-       "      <td id=\"T_279d5_row14_col1\" class=\"data row14 col1\" >Weights_items_features_2</td>\n",
-       "      <td id=\"T_279d5_row14_col2\" class=\"data row14 col2\" >-5.554287</td>\n",
-       "      <td id=\"T_279d5_row14_col3\" class=\"data row14 col3\" >0.245318</td>\n",
-       "      <td id=\"T_279d5_row14_col4\" class=\"data row14 col4\" >-22.641151</td>\n",
-       "      <td id=\"T_279d5_row14_col5\" class=\"data row14 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row14\" class=\"row_heading level0 row14\" >14</th>\n",
+       "      <td id=\"T_ea03d_row14_col0\" class=\"data row14 col0\" >2</td>\n",
+       "      <td id=\"T_ea03d_row14_col1\" class=\"data row14 col1\" >Weights_items_features_2</td>\n",
+       "      <td id=\"T_ea03d_row14_col2\" class=\"data row14 col2\" >-5.554287</td>\n",
+       "      <td id=\"T_ea03d_row14_col3\" class=\"data row14 col3\" >0.245318</td>\n",
+       "      <td id=\"T_ea03d_row14_col4\" class=\"data row14 col4\" >-22.641151</td>\n",
+       "      <td id=\"T_ea03d_row14_col5\" class=\"data row14 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row15\" class=\"row_heading level0 row15\" >15</th>\n",
-       "      <td id=\"T_279d5_row15_col0\" class=\"data row15 col0\" >2</td>\n",
-       "      <td id=\"T_279d5_row15_col1\" class=\"data row15 col1\" >Weights_items_features_3</td>\n",
-       "      <td id=\"T_279d5_row15_col2\" class=\"data row15 col2\" >-13.565555</td>\n",
-       "      <td id=\"T_279d5_row15_col3\" class=\"data row15 col3\" >0.544168</td>\n",
-       "      <td id=\"T_279d5_row15_col4\" class=\"data row15 col4\" >-24.928965</td>\n",
-       "      <td id=\"T_279d5_row15_col5\" class=\"data row15 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row15\" class=\"row_heading level0 row15\" >15</th>\n",
+       "      <td id=\"T_ea03d_row15_col0\" class=\"data row15 col0\" >2</td>\n",
+       "      <td id=\"T_ea03d_row15_col1\" class=\"data row15 col1\" >Weights_items_features_3</td>\n",
+       "      <td id=\"T_ea03d_row15_col2\" class=\"data row15 col2\" >-13.565555</td>\n",
+       "      <td id=\"T_ea03d_row15_col3\" class=\"data row15 col3\" >0.544168</td>\n",
+       "      <td id=\"T_ea03d_row15_col4\" class=\"data row15 col4\" >-24.928965</td>\n",
+       "      <td id=\"T_ea03d_row15_col5\" class=\"data row15 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row16\" class=\"row_heading level0 row16\" >16</th>\n",
-       "      <td id=\"T_279d5_row16_col0\" class=\"data row16 col0\" >2</td>\n",
-       "      <td id=\"T_279d5_row16_col1\" class=\"data row16 col1\" >Weights_items_features_4</td>\n",
-       "      <td id=\"T_279d5_row16_col2\" class=\"data row16 col2\" >-9.794930</td>\n",
-       "      <td id=\"T_279d5_row16_col3\" class=\"data row16 col3\" >0.631004</td>\n",
-       "      <td id=\"T_279d5_row16_col4\" class=\"data row16 col4\" >-15.522781</td>\n",
-       "      <td id=\"T_279d5_row16_col5\" class=\"data row16 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row16\" class=\"row_heading level0 row16\" >16</th>\n",
+       "      <td id=\"T_ea03d_row16_col0\" class=\"data row16 col0\" >2</td>\n",
+       "      <td id=\"T_ea03d_row16_col1\" class=\"data row16 col1\" >Weights_items_features_4</td>\n",
+       "      <td id=\"T_ea03d_row16_col2\" class=\"data row16 col2\" >-9.794930</td>\n",
+       "      <td id=\"T_ea03d_row16_col3\" class=\"data row16 col3\" >0.631004</td>\n",
+       "      <td id=\"T_ea03d_row16_col4\" class=\"data row16 col4\" >-15.522781</td>\n",
+       "      <td id=\"T_ea03d_row16_col5\" class=\"data row16 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "    <tr>\n",
-       "      <th id=\"T_279d5_level0_row17\" class=\"row_heading level0 row17\" >17</th>\n",
-       "      <td id=\"T_279d5_row17_col0\" class=\"data row17 col0\" >2</td>\n",
-       "      <td id=\"T_279d5_row17_col1\" class=\"data row17 col1\" >Weights_items_features_5</td>\n",
-       "      <td id=\"T_279d5_row17_col2\" class=\"data row17 col2\" >-12.126673</td>\n",
-       "      <td id=\"T_279d5_row17_col3\" class=\"data row17 col3\" >0.681118</td>\n",
-       "      <td id=\"T_279d5_row17_col4\" class=\"data row17 col4\" >-17.804060</td>\n",
-       "      <td id=\"T_279d5_row17_col5\" class=\"data row17 col5\" >0.000000</td>\n",
+       "      <th id=\"T_ea03d_level0_row17\" class=\"row_heading level0 row17\" >17</th>\n",
+       "      <td id=\"T_ea03d_row17_col0\" class=\"data row17 col0\" >2</td>\n",
+       "      <td id=\"T_ea03d_row17_col1\" class=\"data row17 col1\" >Weights_items_features_5</td>\n",
+       "      <td id=\"T_ea03d_row17_col2\" class=\"data row17 col2\" >-12.126673</td>\n",
+       "      <td id=\"T_ea03d_row17_col3\" class=\"data row17 col3\" >0.681118</td>\n",
+       "      <td id=\"T_ea03d_row17_col4\" class=\"data row17 col4\" >-17.804060</td>\n",
+       "      <td id=\"T_ea03d_row17_col5\" class=\"data row17 col5\" >0.000000</td>\n",
        "    </tr>\n",
        "  </tbody>\n",
        "</table>\n"
       ],
       "text/plain": [
-       "<pandas.io.formats.style.Styler at 0x7f7d56306970>"
+       "<pandas.io.formats.style.Styler at 0x7fa1a84488b0>"
       ]
      },
      "execution_count": null,
@@ -350,6 +366,156 @@
     "report.style.apply(format_color_groups, axis=None)"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lc_model_2 = LatentClassSimpleMNL(n_latent_classes=3, fit_method=\"EM\", optimizer=\"lbfgs\", epochs=2000, lbfgs_tolerance=1e-6)\n",
+    "hist, results = lc_model_2.fit(elec_dataset, verbose=1)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lc_model_2.latent_logits"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "plt.plot(results)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "break"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lc_model_2.instantiate(\n",
+    "                n_items=elec_dataset.get_n_items(),\n",
+    "                n_shared_features=elec_dataset.get_n_shared_features(),\n",
+    "                n_items_features=elec_dataset.get_n_items_features(),\n",
+    "            )"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "hist_logits = []\n",
+    "hist_loss = []\n",
+    "\n",
+    "# Initialization\n",
+    "init_sample_weight = np.random.rand(3, len(elec_dataset))\n",
+    "init_sample_weight = init_sample_weight / np.sum(init_sample_weight, axis=0, keepdims=True)\n",
+    "for i, model in enumerate(lc_model_2.models):\n",
+    "    # model.instantiate()\n",
+    "    model.fit(elec_dataset, sample_weight=np.clip(init_sample_weight[i], 1e-4, 1), verbose=2)\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lc_model_2.models[2].exact_nll.epsilon"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "init_sample_weight[2].min()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lc_model_2.models[2].predict_probas(elec_dataset)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "tf.math.log(1.0 * 1e-40)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "lc_model_2.models[2]._trainable_weights"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "myw, myl, myloss = [], [], []\n",
+    "\n",
+    "for i in range(10):\n",
+    "    lc_model_2.weights, loss = lc_model_2._expectation(elec_dataset)\n",
+    "    lc_model_2.latent_logits = lc_model_2._maximization(elec_dataset, verbose=2)\n",
+    "\n",
+    "    myw.append(lc_model_2.weights)\n",
+    "    myl.append(lc_model_2.latent_logits)\n",
+    "    myloss.append(loss)\n",
+    "    if np.sum(np.isnan(self.latent_logits)) > 0:\n",
+    "            print(\"Nan in logits\")\n",
+    "            break"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "for i in tqdm.trange(self.epochs):\n",
+    "    self.weights, loss = self._expectation(choice_dataset)\n",
+    "    self.latent_logits = self._maximization(choice_dataset, verbose=verbose)\n",
+    "    hist_logits.append(self.latent_logits)\n",
+    "    hist_loss.append(loss)\n",
+    "    if np.sum(np.isnan(self.latent_logits)) > 0:\n",
+    "        print(\"Nan in logits\")\n",
+    "        break\n",
+    "return hist_logits, hist_loss"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {