EuclideanLoss¶

class dragon.vm.caffe.core.layers.EuclideanLoss(layer_param)[source]¶

Compute the element-wise squared error.

The EuclideanLoss function is defined as:

\[\text{L2Loss}(x, y) = 0.5(x - y)^{2} \]

Examples:

layer {
  type: "EuclideanLoss"
  bottom: "bbox_pred"
  bottom: "bbox_target"
  top: "bbox_loss"
  loss_param {
    normalization: BATCH_SIZE
  }
}