Professor Zhang Xiaoqin from the College of Computer and Artificial Intelligence published a paper entitled “Top-k Feature Selection Framework Using Robust 0-1 Integer Programming” in IEEE T-NNLS (DOI: 10.1109/TNNLS.2020.3009209) as the first author.

The purpose of feature selection is to find the key features in the data while removing redundancy and noise features. It can not only effectively improve the performance of the learning algorithm, but also significantly reduce the computational complexity of the algorithm, which has important value in machine learning. Most feature selection methods use a certain criterion to calculate the scores of all features, then sort the features according to the scores from high to low, and use the K features with the highest scores as the result. Since the local optimal is generally not as good as the global optimal, the set of K good features may not be the best set of K features. In order to solve the above problems, this paper proposes a theoretical framework of feature selection based on 0-1 integer programming approximate optimization algorithm. Under this framework, new feature selection methods can be proposed. Through these methods, an approximately optimal set of K features under a given standard, that is, an approximate global optimal solution can be obtained. Based on this framework, the author respectively proposes an unsupervised feature selection method and a semi-supervised feature selection method. Finally, these two methods are verified on the benchmark public data set, and the results prove the effectiveness of the feature selection method using 0-1 integer programming proposed in this paper. This work was funded by the National Natural Science Foundation of China and the Ministry of Science and Technology.

IEEE T-NNLS, with the full name of IEEE Transactions on Neural Networks and Learning Systems, is a top journal in fields such as artificial intelligence, pattern recognition, image processing, and computer vision. Its Impact Factor in 2018 was 11.683.