Non-convex optimizationIt is used in the fields of machine learning and signal processing, mainly for non-convex problems, that is, it directly solves the problem without using relaxation processing and directly optimizes the non-convex formula.

Common non-convex optimization techniques include the following:

• Projected Gradient Descent
• Alternating Minimization
• Expectation-Maximization Algorithm
• Stochastic Optimization and Its Variants

These methods are fast in practice. Currently, deep learning and some machine learning problems involve non-convex optimization processing.

Transformation for non-convex optimization

• Modify the objective function to transform it into a convex function;
• Discard the constraints and make the new feasible domain a convex set.