1.4 Dissertation Organization
The remainder of this dissertation is organized in the following fashion. In
Chapter 2, we begin with background theory review and analyze their properties.
In Chapter 3, 4, 5, 6 we introduce our major works. In Chapter 3, we present robust
clustering methods with affinity transformation for heterogeneous density clusters
that against scaling parameter tuning and noise sensitivity. Chapter 4 introduces two
unsupervised anomaly detection algorithms: one is a heat-diffusion-based anomaly
detection, another is a quantum-mechanics-based algorithm with strong probabilistic
interpretation. Chapter 5 presents a noise-resistant unsupervised feature selection
algorithm based on multi-perspective correlation measurement. In Chapter 6
we describe a spectral embedding approximation, which is both efficient and effective
for large-scale datasets. Therefore it makes our framework practical in real
world applications. Finally, Chapter 7 summarizes our contributions on the finished
work and also discusses our ongoing work and future research directions.