MLE and MAP - shivamvats/notes GitHub Wiki

Goal: Estimate a parameter a on the basis of a given model of the data (parameterized by a) and some data D.

How: Find a point estimate for a that maximizes the likelihood P(D|a).

As the name suggests, it maximizes the likelihood function of the parameters (given the data). It is a frequentist technique that gives a point estimate of the parameter a, i.e, a = max P(D|a).

• The best explanation of the Likelihood function.

Goal: Estimate a parameter a on the basis of a given model of the data (parameterized by a), some data D and a prior over a.

How: Find a point estimate for a that maximizes the posterior P(a|D).

P(a|D) = P(D|a)P(a).

Here, P(a) is essentially weighting the likelihood for different values of a. So it is just a more general version of MLE.