Multi margin loss
Creates a criterion that optimizes a multi-class classification hinge
loss (margin-based loss) between input x (a 2D mini-batch Tensor) and
output y (which is a 1D tensor of target class indices,
0 ≤q y ≤q \mbox{x.size}(1)-1):
nn_multi_margin_loss(p = 1, margin = 1, weight = NULL, reduction = "mean")
p |
(int, optional): Has a default value of 1. 1 and 2 are the only supported values. |
margin |
(float, optional): Has a default value of 1. |
weight |
(Tensor, optional): a manual rescaling weight given to each
class. If given, it has to be a Tensor of size |
reduction |
(string, optional): Specifies the reduction to apply to the output:
|
For each mini-batch sample, the loss in terms of the 1D input x and scalar output y is:
\mbox{loss}(x, y) = \frac{∑_i \max(0, \mbox{margin} - x[y] + x[i]))^p}{\mbox{x.size}(0)}
where x \in ≤ft\{0, \; \cdots , \; \mbox{x.size}(0) - 1\right\} and i \neq y.
Optionally, you can give non-equal weighting on the classes by passing
a 1D weight tensor into the constructor.
The loss function then becomes:
\mbox{loss}(x, y) = \frac{∑_i \max(0, w[y] * (\mbox{margin} - x[y] + x[i]))^p)}{\mbox{x.size}(0)}
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.