Loss Functions

Cross Entropy Loss

Calculates a score that summarizes correctnes of set of classifications. Value close to 0 is better, high values are worse.

The target value should be the index of the correct class. Each vector location should be a probability of the input belonging to the corresponding index.

Example

classification: [.0001, .988]
target: 1

Pytorch take a batches of classification.

classifications:
  [
    [.0001, .988],
    [.923, .0001],
    [.923, .0003],
  ]
targets:
  [1, 0, 0]

Typical input is a vector of probabilities, where each element represents the probability of the input belonging to a particular class. The target is a vector of probabilities, where each element is a an integer label representing the class.

See more on:

• Geek for Geeks - Cross Entropy