A two-layer model for music pleasure regression
Xing Wang, Yuqian Wu, Xiaoou Chen, Deshun Yang · 2013
We adopt a two-layer regression model for music pleasure regression. Pleasure orientation of a song is estimated first, and then different regressors are used to predict degree of pleasure according to the estimated orientation. By using corresponding regressors for each instance, there is a big improvement when we assume the first layer is perfect in comparison with one-layer model. By tuning the confidence threshold of the orientation classifier of the two-layer model, we get improvement over one-layer model.