3VL-MLP: A way for realizing interpretable aware systems
Qiangfu Zhao · 2015
The goal of awareness computing (AC) is to establish a win-win relation between service systems and their users. If a system is aware of the user needs and the proper timing for providing proper services, the system will be popular and profitable. In many cases, it is also necessary to tell the users reasons why certain services are provided at a certain time point. The ability of providing understandable reasons can make the system more trustable, and thus more popular and profitable. We believe that designing interpretable aware systems is extremely important to achieve the goal of AC. In this paper, we first provide a formal definition of aware systems, and then consider a way to build interpretable aware systems based on 3-valued logic. As a concrete example, we use a 3-valued logic multilayer perceptron (3VL-MLP) to show a way for realizing interpretable aware systems. Results on several public databases show that the 3VL-MLPs perform relatively well, and can be useful in practice.