In a supervised learning framework, we have a dataset $D=\{(x_1,y_1),(x_2,y_2),$dots,$(x_n,y_n)\}.$ Prove$/$Disprove that minimization of cross$-$entropy loss with one$-$hot encoding is a special case of solving the maximum likelihood estimation.