Introduced on day 32
Logit
A network's raw, unnormalized output score for a class, any real number; a sigmoid or softmax turns logits into probabilities.
In detail
In logistic regression the logit is the log-odds. Training losses such as cross-entropy are best computed from logits with a log-sum-exp trick, which is numerically stable and avoids applying softmax twice. A logit is not a probability and should not be thresholded at 0.5 as if it were.