Feature Selection: Hybrid or Embedded Approaches

Question

I do research on the selection of functions, and I do not understand the difference in these two approaches.

According to most authors in the literature, function selection algorithms are divided into three categories. The first two, the filter and the wrapper are easy to understand, and there is general agreement on this. However, there seems to be a misunderstanding in the latter category. Some authors as the case of H. Liu called the latter category hybrid. In contrast, V. Kumar calls it inline. Depending on this, there are cases when authors define 4 categories, including both built-in and hybrid algorithms, as in the case of P. Abinaya .

The authors explain hybrid algorithms as a combination between a filtering algorithm and wrapper approaches. The main idea of ​​these algorithms is to use a filter approach to reduce the search space for the wrapper approach.

On the other hand, the definition of embedded algorithms in the literature is very dependent on the source. Some use almost the same certainty as hybrid algorithms, as with the wikipedia page . Others provide more abstract definitions, such as methods that perform function selection when studying optimal parameters , and methods that include knowledge of the specific structure of the class of functions used by a particular learning machine .

Therefore, I would appreciate it if someone could explain to me what the difference is between the two approaches or to give a less abstract definition of built-in methods.

Thank.