[source]

Least Squares

Least Squares (or quadratic loss) is a function that computes the average squared error (MSE) between the target output given by the labels and the actual output of the network. It produces a smooth bowl-shaped gradient that is highly-influenced by large errors.

\[ Least Squares = \sum_{i=1}^{D}(y_i-\hat{y}_i)^2 \]

Parameters

This cost function does not have any parameters.

Example

use Rubix\ML\NeuralNet\CostFunctions\LeastSquares;

$costFunction = new LeastSquares();