This approach optimizes feature selection using a hard cardinality constraint in nonlinear SVMs, suggesting improvements through metaheuristic and submodular optimization.
Key Points
The proposed method guarantees strict control over the number of selected features in nonlinear SVMs.
The approach includes a local search metaheuristic, which is effective for general nonlinear kernels.
The decomposition framework alternates optimization between continuous and binary variables for enhanced efficiency.
Numerical experiments reveal that the algorithms outperform standard methods for solving mixed-integer nonlinear programming problems.