Concentration Distribution Learning from Label Distributions

Label distribution learning (LDL) is an effective method to predict the relative label description degree (a.k.a. label distribution) of a sample. However, the label distribution is not a complete representation of an instance because it overlooks the absolute intensity of each label. Specifically, it's impossible to obtain the total description degree of hidden labels that not in the label space, which leads to the loss of information and confusion in instances. To solve the above problem, we come up with a new concept named background concentration to serve as the absolute description degree term of the label distribution and introduce it into the LDL process, forming the improved paradigm of concentration distribution learning. Moreover, we propose a novel model by probabilistic methods and neural networks to learn label distributions and background concentrations from existing LDL datasets. Extensive experiments prove that the proposed approach is able to extract background concentrations from label distributions while producing more accurate prediction results than the state-of-the-art LDL methods. The code is available in https://github.com/seutjw/CDL-LD.

Every image has its background, and we always focus on the main components instead. In machine learning, every instances also has its background, and the label only describes its main parts. As a result, instances with the same proportion of main parts and different backgrounds will share an identical label, which is very unreasonable.To this end, we take the proportion of backgrounds in instances into the learning objective, and come up with a new paradigm, which learns labels and background proportions of instances simultaneously. We also design a corresponding learning algorithm, and prove its effectiveness in both theoretical and experimental aspects.For the convenience of subsequent researchers on this idea, we construct a real-world dataset for this paradigm. This further enhances the practical value of our research, making it a promising research direction.