 |
Center for Nonlinear Studies |
We present two graph-based algorithms for classification of high-dimensional data. The algorithms generalize a binary diffuse interface model to the case of multiple classes. Motivated by total variation techniques, the algorithms involve minimizing the Ginzburg-Landau (GL) energy functional adapted to a semi-supervised graph setup. We develop two multiclass generalizations, one based on a scalar representation and other based on a vector-field representation. We compare the performance of the two multiclass formulations in synthetic data as well as real benchmark sets, and demonstrate that our experimental results are competitive with the state-of-the-art among other graph-based algorithms. The talk is based on joint work with Arjuna Flenner, Allon Percus, Ekaterina Merkurjev and Andrea Bertozzi.
Host: Marian Anghel, manghel@lanl.gov