theano 编程技巧

判断并输出

self.output = (lin_out if activation is None else activation(lin_out))

误分率

T.mean(T.neg(y_pred, y))