rl_equation_solver.utilities.loss.LossMixin

class LossMixin

Bases: object

Mixin class with collection of loss functions

Methods

huber_loss(x, y[, delta])	Huber loss.
l2_loss(x, y)	L2 Loss
smooth_l1_loss(x, y)	Smooth L1 Loss

huber_loss(x, y, delta=1.0)

Huber loss. Huber loss, also known as Smooth Mean Absolute Error, is a loss function used in various machine learning and optimization problems, particularly in regression tasks. It combines the properties of both Mean Squared Error (MSE) and Mean Absolute Error (MAE) loss functions, providing a balance between the two.

\[ L(y, f(x)) = \begin{cases} \begin{split} \frac{1}{2} (y - f(x))^2, & \text{ if } |y - f(x)| \leq \delta \\ \delta |y - f(x)| - \frac{1}{2} \delta^2, & \text{ otherwise} \end{split} \end{cases} \]

smooth_l1_loss(x, y)

Smooth L1 Loss

l2_loss(x, y)

L2 Loss