Parameters Tuning

This page contains parameters tuning guides for different scenarios.

For Better Accuracy

• Use large n_nodes_H and n_groups_Z (may be slower)

• Use bigger training data

Deal with Over-fitting

• Use more k_neighbors for semi-supervised learning

• Use bigger training data

• Try reg_alpha, reg_lambda, reg_laplacian and sigma for regularization

• Try n_nodes_Z for generating suitable feature nodes