filters update

lib/errors/_index.md

@@ -34,3 +34,4 @@ Error metrics and loss functions for model/strategy evaluation. All error indica
 | [TUKEY](tukey/Tukey.md) | Tukey Biweight Loss | Hard-rejects outliers beyond threshold. Redescending. |
 | [WMAPE](wmape/Wmape.md) | Weighted MAPE | Volume-weighted percentage error. For heterogeneous data. |
 | [WRMSE](wrmse/Wrmse.md) | Weighted RMSE | Weighted root mean squared error. Custom observation weighting. |
+| STDERR | Standard Error | StdDev / √n. Measures sampling variability. |

lib/errors/logcosh/logcosh.pine

@@ -10,7 +10,7 @@ stable_logcosh(float x) =>
     float LOG2 = 0.6931471805599453
     float absX = math.abs(x)
     // For large values, use asymptotic approximation to avoid overflow
-    absX > 20.0 ? absX - LOG2 : math.log(math.cosh(x))
+    absX > 20.0 ? absX - LOG2 : math.log((math.exp(x) + math.exp(-x)) / 2.0)
 
 //@function Calculates Log-Cosh Loss
 //@param   actual Series of actual values