Publications

1. 2010, Nicola Rebagliati, Alessandro Verri, "A spectral technique for clustering", Neural Computation, submitted [Informal Summary: same as below, but with proofs and a practical example]
2. 2010, Nicola Rebagliati, Alessandro Verri, "A randomized algorithm for spectral clustering", ESANN 2010, Slides, paper
   [Informal Summary: The problem is that of minimizing the normalized cut of a graph with K classes. This is a short paper where I sum up some of my results: first it is a bound for choosing more than K normalized laplacian eigenvectors, second we can use them by sampling random K-dimensional linear subspaces and then round them. Experiments on a benchmark, the graph partitioning archive, show that this idea works.]

PhD Thesis

Master Thesis (Italian)