Curious Feature Selection

In state-of-the-art big-data applications, the process of building machine learning models can be very challenging due to continuous changes in data structures and the need for human interaction to tune the variables and models over time. Hence, expedited learning in rapidly changing environments is required. In this work, we address this challenge by implementing concepts from the field of intrinsically motivated computational learning, also known as artificial curiosity (AC). In AC, an autonomous agent acts to optimize its learning about itself and its environment by receiving internal rewards based on prediction errors. We present a novel method of intrinsically motivated learning, based on the curiosity loop, to learn the data structures in large and varied datasets. An autonomous agent learns to select a subset of relevant features in the data, i.e., feature selection, to be used later for model construction. The agent optimizes its learning about the data structure over time without requiring external supervision. We show that our method, called the Curious Feature Selection (CFS) algorithm, positively impacts the accuracy of learning models on three public datasets.