transition[4][4]

emissions[4][2]

p[2]

alpha(i, 1) = p[i] * emissions[i][o_1]
alpha(i, t) = emissions[i][o_t] * max_{k in states} (alpha(k, t-1) * transition[k][i])

Best_K_Values(t, i) = Top K over all i,preceding_state,k (emissions[i][o_t] * m[preceding_state][k] * transition[preceding_state][i])