• Multi-Perspective Feature Selection
Feature importance is usually more about a “local” conception than a “global”
one. To obtain a better representative feature subsets, the feature impact to
different low-embeddings or spectrums need to be considered [35]. Besides,
• Efficient Approximation of Spectral Embeddings
One foundation of our research framework is spectral embedding construction,
which is one of the most effective dimension reduction algorithms in
machine learning and data mining [100]. However, its associated high complexity
in both time O(n
3
) and space O(n
2
) prevents it from practical utilization
in many large-scale real-world applications. Many researches have
developed a few approximate spectral embeddings which are more efficient,