Abstract:We consider stochastic optimization problems involving an expected value of a nonlinear function of a base random vector and a conditional expectation of another function depending on the base random vector, a dependent random vector, and the decision variables. We call such problems conditional stochastic optimization problems. They arise in many applications, such as uplift modeling, reinforcement learning, and contextual optimization. We propose a specialized single time-scale stochastic method for nonconvex constrained conditional stochastic optimization problems with a Lipschitz smooth outer function and a generalized differentiable inner function. In the method, we approximate the inner conditional expectation with a rich parametric model whose mean squared error satisfies a stochastic version of a Łojasiewicz condition. The model is used by an inner learning algorithm. The main feature of our approach is that unbiased stochastic estimates of the directions used by the method can be generated with one observation from the joint distribution per iteration, which makes it applicable to real-time learning. The directions, however, are not gradients or subgradients of any overall objective function. We prove the convergence of the method with probability one, using the method of differential inclusions and a specially designed Lyapunov function, involving a stochastic generalization of the Bregman distance. Finally, a numerical illustration demonstrates the viability of our approach.

Heuristic Parameter Selection Based on Functional Minimization: Optimality and Model Function Approach

A New Algorithm For Unconstrained Optimization Problem With The Form Of Sum Of Squares Minimization

Heuristic parameter-choice rules for convex variational regularization based on error estimates

A Kernel Function Optimization And Selection Algorithm Based On Cost Function Maximization

A Variance Minimization Criterion to Feature Selection Using Laplacian Regularization

A discretizing Tikhonov regularization method via modified parameter choice rules

Model Functions in the Modified L-Curve Method-Case Study: the Heat Flux Reconstruction in Pool Boiling

Model Misfit Minimization

Discrepancy based model selection in statistical inverse problems

A Unified Dynamic Approach to Sparse Model Selection

Conservative parametric optimality and the ridge method for tame min-max problems

A Functional Model Method for Nonconvex Nonsmooth Conditional Stochastic Optimization

Adaptive choice of the Tikhonov regularization parameter to solve ill-posed linear algebraic equations via Liapunov Optimizing Control

Gradient-based Regularization Parameter Selection for Problems with Non-smooth Penalty Functions

A constrained minimum criterion for model selection

Surface Approximation Via Sparse Representation and Parameterization Optimization

A convex dual problem for the rational minimax approximation and Lawson's iteration

Understanding Implicit Regularization in Over-Parameterized Single Index Model

A Gradient Descent Based Algorithm for ℓ Minimization.

Parameter choice strategies for regularized least squares approximation of noisy continuous functions on the unit circle

Multi-parameter Regularization and Its Numerical Realization