MultiHingeLossLayer.md

CMultiHingeLossLayer Class

• CMultiHingeLossLayer Class
  • Settings
    • Loss weight
    • Gradient clipping
  • Trainable parameters
  • Inputs
  • Outputs
    • Getting the value of the loss function

This class implements a layer that calculates a hinge loss function for a classification scenario with multiple classes.

The function is calculated according to the formula:

loss = max(0, 1 - (x_right - x_max_wrong))

where:

• x_right is the network response for the class to which the object belongs.
• x_max_wrong is the largest of the responses for the other classes.

Settings

Loss weight

void SetLossWeight( float lossWeight );

Sets the multiplier for this function gradient during training. The default value is 1. You may wish to change the default if you are using several loss functions in your network.

Gradient clipping

void SetMaxGradientValue( float maxValue );

Sets the upper limit for the absolute value of the function gradient. Whenever the gradient exceeds this limit its absolute value will be reduced to GetMaxGradientValue().

Trainable parameters

This layer has no trainable parameters.

Inputs

The layer may have 2 to 3 inputs:

1. The network output for which you are calculating the loss function.

2. The class labels represented by a blob in one of the two formats:

   • the blob contains float data, with BatchLength, BatchWidth, and ListSize equal to these dimensions of the first input, and GetObjectSize() equal to the number of classes. It should contain the probability distribution for an object to belong to any of those classes.
   • the blob contains int data with BatchLength, BatchWidth, and ListSize equal to these dimensions of the first input, and the other dimensions equal to 1. Each object in the blob contains the number of the class to which the corresponding object from the first input belongs.

3. [Optional] The objects' weights. This input should have the same BatchLength, BatchWidth, and ListSize dimensions as the first input. Height, Width, Depth, and Channels should be equal to 1.

Outputs

This layer has no output.

Getting the value of the loss function

float GetLastLoss() const;

Use this method to get the value of the loss function calculated on the network's last run.