The field of optimization and modeling is rapidly advancing with new methods and techniques being developed to improve the efficiency and accuracy of various models and algorithms. One of the key areas of focus is on developing more effective optimization algorithms that can handle complex and high-dimensional problems. This includes the use of bi-fidelity approaches, which leverage a combination of low-fidelity and high-fidelity function evaluations to improve the efficiency of optimization. Additionally, there is a growing interest in developing more robust and reliable models that can handle uncertainty and model discrepancy. This includes the use of probabilistic approaches, such as Bayesian optimization and Bayesian experimental design, which can provide a more comprehensive understanding of the underlying systems and processes. Overall, the field is moving towards more sophisticated and powerful methods that can handle complex problems and provide more accurate and reliable results. Notable papers in this area include the work on FigBO, a generalized acquisition function framework with look-ahead capability for Bayesian optimization, and the development of DiffLiB, a high-fidelity differentiable modeling framework for lithium-ion batteries. These papers demonstrate the potential of these new methods and techniques to advance the field and provide more effective solutions to complex problems.
Advances in Optimization and Modeling
Sources
Improving the Threshold for Finding Rank-1 Matrices in a Subspace
Hyper-differential sensitivity analysis with respect to model discrepancy: Prior Distributions
Stochastic Subspace via Probabilistic Principal Component Analysis for Characterizing Model Error
Low-Rank Matrix Approximation for Neural Network Compression
FigBO: A Generalized Acquisition Function Framework with Look-Ahead Capability for Bayesian Optimization
Bayesian Experimental Design for Model Discrepancy Calibration: An Auto-Differentiable Ensemble Kalman Inversion Approach
DiffLiB: High-fidelity differentiable modeling of lithium-ion batteries and efficient gradient-based parameter identification
Non-Linear Modeling and Analysis of Amplifier-Less Potentiostat Architectures
Bayesian Optimization-based Tire Parameter and Uncertainty Estimation for Real-World Data
Guessing Efficiently for Constrained Subspace Approximation
On Advancements of the Forward-Forward Algorithm
Stochastic Subspace Descent Accelerated via Bi-fidelity Line Search
The Planted Orthogonal Vectors Problem
Error bounds for function approximation using generated sets
Subspace-Distance-Enabled Active Learning for Efficient Data-Driven Model Reduction of Parametric Dynamical Systems