Loss Functions

Feature −[ylogy^​+(1−y)log(1−y^​)] −log(y^​)
Typical Task Binary Classification (Yes/No) Multi-class (Cat, Dog, Bird)
Last Layer Sigmoid (one output neuron) Softmax (multiple output neurons)
Logic "How far are you from the 1 AND the 0?" "How much probability did you give the right answer?"
Pasted image 20260426225757.png