Embedded methods are a type of feature selection technique in machine learning that perform feature selection as part of the model construction process. They are often more efficient than other methods because they integrate feature selection with model training, allowing for the simultaneous optimization of both tasks.