Use ElasticNet's regularization path generator when using l2 loss

regularization.py

@@ -17,8 +17,7 @@ from sklearn.svm import LinearSVC
 from sklearn.datasets import load_iris
 from sklearn.metrics import classification_report, mean_squared_error, mean_absolute_error
 from sklearn.datasets import make_regression
-from sklearn.linear_model import Ridge
-from sklearn.linear_model import Lasso
+from sklearn.linear_model import ElasticNet
 
 import traceback
 import yaml
@@ -65,6 +64,20 @@ def l2_loss(W, y, X):
     return np.mean((y - preds) ** 2, axis=1).reshape(num_dots, num_dots)
 
 
+def l2_loss_regularization_path(y, X, regularization_type):
+    if regularization_type == "l2":
+        l1_ratio = 0
+        alphas = np.arange(0, 100, 0.02)
+    elif regularization_type == "l1":
+        l1_ratio = 1
+        alphas = None
+    else:
+        raise ValueError("regularization_type must be 'l1' or 'l2'")
+
+    _, coefs, *_ = ElasticNet.path(X, y, l1_ratio=l1_ratio, alphas=alphas)
+    return coefs.T
+
+
 class Regularization:
     def __init__(self, width, height):
         # initialized in draw_plot
@@ -211,17 +224,21 @@ class Regularization:
 
         # plot path
         if self.plot_regularization_path:
-            min_loss_reg = regs.ravel()[np.argmin(losses)]
-            path_reg_levels = np.linspace(0, min_loss_reg, 20)
-            path_w = []
-            for reg_level in path_reg_levels:
-                mask = regs <= reg_level
-                if np.sum(mask) == 0:
-                    continue
-                idx = np.argmin(losses[mask])
-                path_w.append(stacked[mask][idx])
-            
-            path_w = np.array(path_w)
+            if self.loss_type == "l2":
+                path_w = l2_loss_regularization_path(y, X, regularization_type=self.reg_type)
+            else:
+                min_loss_reg = regs.ravel()[np.argmin(losses)]
+                path_reg_levels = np.linspace(0, min_loss_reg, 20)
+                path_w = []
+                for reg_level in path_reg_levels:
+                    mask = regs <= reg_level
+                    if np.sum(mask) == 0:
+                        continue
+                    idx = np.argmin(losses[mask])
+                    path_w.append(stacked[mask][idx])
+                
+                path_w = np.array(path_w)
+
             ax.plot(path_w[:, 0], path_w[:, 1], "r-")
 
         # custom legend
@@ -333,7 +350,7 @@ class Regularization:
 
                         with gr.Row():
                             # regularizer type 
-                            regularizer_type = gr.Dropdown(choices=['l1', 'l2', 'elastic-net'],
+                            regularizer_type = gr.Dropdown(choices=['l1', 'l2'],
                                                            label='Regularizer type',
                                                            value='l2',
                                                            visible=True)