Why to take logarithm in likelihood estimate

1. Often assume that the characteristics of the independent distribution of the same, then the likelihood function is often the form of liancheng, directly to cheat to the bad, according to log can be connected to add.

2. Another probability value is a decimal, multiple decimals multiplied easily zhao floating point overflow, go to log into a plus can avoid this problem.

If there is no connection in the original likelihood function, then go to the post-operative no log (a+b) Form, at this time can use GD, otherwise with EM, Sonjuku personal understanding.

Why to take logarithm in likelihood estimate