Hello! I have question regarding statistical models with latent variables. Thank you in advance!

(a) Explain why is it generally necessary to marginalize over the latent variables - i.e. p(X|0) = Ezp(X, Z|0). (b) Given this marginalization, explain why is it often difficult to compute the maximum likelihood solution for mod- els with latent variables. (c) Explain how EM overcomes the mathematical difficulty described in (b) when computing maximum likelihood solutions