If a likelihood function has many local maximum points, then depending on the initial values, the Maximum Likelihood Estimation (MLE) may get stuck in the local maximum points.

Now, if the likelihood function has too many local minimum and maximum (too many that it gets close to a periodic function), guessing an initial value for the optimization process may be very difficult, as the estimation process is likely to get stuck in local maxima very frequently.

My question is how are such likelihood functions (almost periodic functions) estimated?