Fix rss calculation (#7)

* Fix model selection

* Run example clean

* Normalize rss using mean weights in gamlss update

* Update mean_of_weights in update

* Fix init of mean_of_weights_inner object

* Fix rss calculation

* Bump version number

---------

Co-authored-by: simon-hirsch <simon.hirsch@stud.uni-due.de>

pyproject.toml

@@ -6,7 +6,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "rolch"
-version = "0.1.3"
+version = "0.1.4"
 authors = [
   {name="Simon Hirsch", email="simon.hirsch@stud.uni-due.de"},
   {name="Jonathan Berrisch", email="jonathan.berrisch@uni-due.de"},

src/rolch/online_gamlss.py

@@ -239,7 +239,7 @@ def update_beta(
 
             rss = (
                 (residuals**2).flatten() * w
-                + (1 - self.forget) * (self.rss[param] * self.sum_of_weights[param])
+                + (1 - self.forget) * (self.rss[param] * self.mean_of_weights[param])
             ) / (self.mean_of_weights[param] * (1 - self.forget) + w)
 
         elif (self.method == "lasso") & self.intercept_only[param]:
@@ -263,7 +263,8 @@ def update_beta(
 
             rss = (
                 (residuals**2).flatten() * w
-                + (1 - self.forget) * (self.rss[param] * self.sum_of_weights[param])
+
+                + (1 - self.forget) * (self.rss[param] * self.mean_of_weights[param])
             ) / (self.mean_of_weights[param] * (1 - self.forget) + w)
 
         elif self.method == "lasso":
@@ -292,7 +293,8 @@ def update_beta(
 
             rss = (
                 (residuals**2).flatten() * w
-                + (1 - self.forget) * (self.rss[param] * self.sum_of_weights[param])
+
+                + (1 - self.forget) * (self.rss[param] * self.mean_of_weights[param])
             ) / (self.mean_of_weights[param] * (1 - self.forget) + w)
 
             model_params_n = np.sum(np.isclose(beta_path, 0), axis=1)
@@ -510,6 +512,7 @@ def update(
         self.y_gram_inner = copy.copy(self.y_gram)
         self.rss_inner = copy.copy(self.rss)
         self.sum_of_weights_inner = copy.copy(self.sum_of_weights)
+        self.mean_of_weights_inner = copy.copy(self.mean_of_weights)
 
         self.lambda_max_inner = copy.copy(self.lambda_max)
         self.lambda_path_inner = copy.copy(self.lambda_path)
@@ -536,6 +539,7 @@ def update(
         self.x_gram = copy.copy(self.x_gram_inner)
         self.y_gram = copy.copy(self.y_gram_inner)
         self.sum_of_weights = copy.copy(self.sum_of_weights_inner)
+        self.mean_of_weights = copy.copy(self.mean_of_weights_inner)
         self.rss = copy.copy(self.rss_inner)
 
         self.lambda_max = copy.copy(self.lambda_max_inner)
@@ -880,6 +884,9 @@ def _inner_update(
             self.sum_of_weights_inner[param] = (
                 np.sum(w * wt) + (1 - self.forget) * self.sum_of_weights[param]
             )
+            self.mean_of_weights_inner[param] = (
+                self.sum_of_weights_inner[param] / self.n_training
+            )
 
             olddv = dv