Huber Loss

Reference resources    Huber Loss - cloud + Community - Tencent cloud

Huber Loss Is a loss function with parameters for regression problems , The advantage is that it can enhance the square error loss function (MSE, mean square error) Robustness to outliers .

When the prediction deviation is less than δ when , It uses squared error , When the prediction deviation is greater than δ when , Linear error used .

Compared with the linear regression of the least squares ,HuberLoss Reduced penalties for outliers , therefore HuberLoss It is a commonly used robust regression loss function .

Huber Loss The definition is as follows

                            

Parameters a Usually means residuals, writing y−f(x), When a = y−f(x) when ,Huber loss Defined as ：

                 

δ yes HuberLoss Parameters of ,y Is the real value ,f(x) Is the predicted value of the model , And by definition Huber Loss You can lead everywhere .