Merge pull request #66 from Gromwud/main

Cumulitive update

Author: Gromwud

epde/globals.py

@@ -104,7 +104,7 @@ class VerboseManager:
 def init_verbose(plot_DE_solutions : bool = False, show_iter_idx : bool = True,
                  show_iter_fitness : bool = False, show_iter_stats : bool = False,
                  show_ann_loss : bool = False, show_warnings : bool = False,
-                 candidate_objectives : bool = False):
+                 candidate_objectives : bool = True):
     """
     Method for initialized of manager for output in text form
 

epde/operators/common/sparsity.py

@@ -55,12 +55,15 @@ def fit(self, X, y, sample_weights):
         residual = y - (X @ self.coef_ + self.intercept_)
 
         iteration = 0
-        max_change = np.inf
 
         # 2. Coordinate Descent Loop
         while iteration < self.max_iter and not all(self.coef_ == 0):
+            max_change = 0
+            max_abs_coef = 0.0
+
             # Sort features by instability (highest CV first)
-            for j in np.argsort(cv)[::-1]:
+            indices = np.argsort(cv)[::-1]
+            for j in indices:
                 old_coef = self.coef_[j]
 
                 if old_coef == 0:
@@ -91,15 +94,57 @@ def fit(self, X, y, sample_weights):
                     self.intercept_ = weights.mean(axis=0)[-1]
                     residual = y - (X @ self.coef_ + self.intercept_)
                     iteration = 0
+                    max_change = np.inf
                     break
 
                 residual -= (new_coef - old_coef) * X[:, j]
-                change = abs(new_coef - old_coef) / abs(old_coef)
+                change = abs(new_coef - old_coef)
+                if change > max_change:
+                    max_change = change
+                # change = abs(new_coef - old_coef) / abs(old_coef)
                 # change = abs(self.intercept_ - old_intercept) / abs(old_intercept)
-                max_change = max(max_change, change)
+                # max_change = max(max_change, change)
+
+            max_abs_coef = np.max(np.abs(self.coef_))
+
+            # Критерий 1: max_j |w_new - w_old| <= tol * max_j |w_j|
+            if max_change <= self.tol * max_abs_coef:
+                # Критерий 2: Dual Gap <= tol * ||y||^2 / n_samples
+                # Вычисляем компоненты дуального зазора
+                # Примечание: Для Lasso с весами lambda_j = threshold_j
+
+                # 1. Вычисляем корреляции признаков с остатками
+                xt_residual = X.T @ residual
+                y_sq_sum = np.sum((y - self.intercept_) ** 2)
+
+                # 2. Масштабирующий фактор для обеспечения дуальной допустимости
+                # В sklearn: dual_scale = min(1, alpha / max(|X.T @ res|))
+                # Здесь используем ваши индивидуальные threshold_j
+                dual_norm = 0
+                for j in range(self.n_features):
+                    if cv[j] * y_sq_sum > 0:
+                        dual_norm = max(dual_norm, abs(xt_residual[j]) / cv[j] * y_sq_sum)
+
+                if dual_norm > 1.0:
+                    const_residual = residual / dual_norm
+                else:
+                    const_residual = residual
+
+                # 3. Вычисление Gap: Primal Objective - Dual Objective
+                # Primal = 0.5 * ||res||^2 + sum(threshold_j * |w_j|)
+                # Dual = 0.5 * ||y-intercept||^2 - 0.5 * ||y-intercept - const_residual||^2
+                primal_obj = 0.5 * np.sum(residual ** 2) + np.sum(cv * y_sq_sum * np.abs(self.coef_))
+                dual_obj = 0.5 * y_sq_sum - 0.5 * np.sum((y - self.intercept_ - const_residual) ** 2)
+
+                dual_gap = primal_obj - dual_obj
+
+                # Итоговая проверка по формуле со скрина
+                if dual_gap <= self.tol * (y_sq_sum / self.n_samples):
+                    break
+
+            # if max_change < self.tol:
+            #     break
 
-            if max_change < self.tol:
-                break
             iteration += 1
         # print(iteration)
         return self