“Highway” and “residual” constructs scale better by sometimes using the identity function as one of these functions. A “self-gating” one of these functions named Swish is atypically non-monotonic. These functions often have overall worse performance when their slope is near zero (15[1])in so-called (15[1])saturating regions; an issue (15[1])in which one of these functions can become almost completely inactive in a saturating region is resolved by a (*) “leaky” variant. The softmax is used (10[1])as the (10[1])final one of these functions to “select” (-5[1])between classes in classification problems. (10[1])The hyperbolic tangent and sigmoids have largely been replaced as the current standard (-5[1])for these functions (10[1])by a simple piecewise-defined “rectified linear” one named ReLU (“REL-yoo”). For 10 points, name these functions applied to a weighted sum to produce the output of each node in a neural network. ■END■ (0[1])