Lines Matching full:loss
23 // SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 // For the exact manner in which loss functions are converted to the
37 // squared loss is expected, it is not necessary to create a loss
39 // residuals implies a standard squared loss.
43 // measurements, it is important to use a loss function that reduces
57 // anything special (i.e. if we used a basic quadratic loss), the
59 // due to the quadratic nature of squared loss. This results in the
64 // Using a robust loss function, the cost for large residuals is
70 // In general, there isn't a principled way to select a robust loss
72 // only experimenting with robust loss functions if standard squared
73 // loss doesn't work.
91 // is required to fill in the value and derivatives of the loss
130 // loss function when building the problem accomplishes the same
220 // Loss that is capped beyond a certain level using the arc-tangent function.
222 // For costs much smaller than 'a', the loss function is linear and behaves like
240 // Loss function that maps to approximately zero cost in a range around the
277 // Composition of two loss functions. The error is the result of first
279 // The loss functions must not be NULL.
305 // Since we treat the a NULL Loss function as the Identity loss
311 // Constructs a ScaledLoss wrapping another loss function. Takes
312 // ownership of the wrapped loss function or not depending on the
332 // wish to mutate the scale of the loss function. For example, when
336 // better convergence behaviour than just using a loss function with a
339 // This templated class allows the user to implement a loss function
