Semi-supervised Local Linear Embed Algorithm Based on Side-information

The local linear embedded algorithm (LLE) is a typical no-supervised learning method. Due to that LLE cannot utilize the known information, it cannot achieve a perfect learning result. Side-information is also a kind of label information, which is relatively easier to get. Based on it, a new algorithm, semi-supervised local linear embedded (SFLLE), has been developed. This new method utilize the both the positive and negative constrain to overcome the shortcoming of the previous LLE method. The experiment result demonstrated the validation of this method.