I have read some articles which mention "Quadratic" error function several times. I know the form and plot of the error function, but not sure about "Quadratic" error function.
Could someone, please explain it to me.
Generally, it means proportional to the square of something. Traditional regression, for example, is also known as OLS, or Ordinary Least Squares, which is a quadratic loss function.
Quadratic loss functions have a natural value, since the central limit theorem makes many random models converge on a normal distribution, an then the only two parameters are the mean and the variance, the latter of which is quadratic.
Usually, when you see the term quadratic, it is referring to – or allowing – generalization to more than one dimension. if $x$ is a vector, and $A$ is a matrix, and they conform, then $x'Ax$ is a generalized 'quadratic form'.
- Solved – Loss function (and encoding?) for angles
- Solved – No linear effects but a quadratic effect
- Solved – Why is the default cost function choice of a neuron quadratic loss
- Solved – Why does differencing once remove not only linear but also nonlinear trends
- Solved – the loss function of hard margin SVM