Dynamic Graph Learning for Feature Selection

In the era of big data, data presents multi-view, high-dimensional and complex characteristics. For one thing, with multi-view features, the data could be characterized more precisely and comprehensively from different perspectives. For another, high-dimensional multi-view features inevitably generate expensive computation costs and cause massive storage costs. Moreover, since raw data generally contains adverse noise, outlying entries, irrelevant and redundant features, the intrinsic dimension of the data may be much lower than the dimension of the raw data. The low-dimensional and robust data representation can effectively improve model efficiency as well as accuracy and thus is vital for downstream tasks in various fields.