global convergence
Global convergence means that an optimization algorithm is guaranteed to find a global minimum of the loss function, regardless of the initialization, which is a desirable characteristic for training deep learning models.
- A Computationally Viable Numerical Gradient-based Technique for Optimal Covering Problems
- A Regularized Newton Method for Nonconvex Optimization with Global and Local Complexity Guarantees
- Affine-Invariant Global Non-Asymptotic Convergence Analysis of BFGS under Self-Concordance
- Block Coordinate Descent for Neural Networks Provably Finds Global Minima
- Convergence of the Gradient Flow for Shallow ReLU Networks on Weakly Interacting Data
- Global Convergence for Average Reward Constrained MDPs with Primal-Dual Actor Critic Algorithm
- Guarantees for Alternating Least Squares in Overparameterized Tensor Decompositions
- Large Stepsizes Accelerate Gradient Descent for Regularized Logistic Regression
- Learning Cocoercive Conservative Denoisers via Helmholtz Decomposition for Poisson Imaging Inverse Problems
- On the Convergence of Single-Timescale Actor-Critic
- Preconditioned Langevin Dynamics with Score-based Generative Models for Infinite-Dimensional Linear Bayesian Inverse Problems
- Solving Neural Min-Max Games: The Role of Architecture, Initialization & Dynamics
- Targeted Maximum Likelihood Learning: An Optimization Perspective