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.
Parameters#
This cost function does not have any parameters.
Example#
use Rubix\ML\NeuralNet\CostFunctions\LeastSquares;
$costFunction = new LeastSquares();
Last update:
2021-01-25