Iterative surrogate model optimization (ISMO): An active learning algorithm for PDE constrained optimization with deep neural networks
نویسندگان
چکیده
We present a novel active learning algorithm, termed as iterative surrogate model optimization (ISMO), for robust and efficient numerical approximation of PDE constrained problems. This algorithm is based on deep neural networks its key feature the selection training data through feedback loop between any underlying standard algorithm. Under suitable hypotheses, we show that resulting optimizers converge exponentially fast (and with decaying variance), respect to increasing number samples. Numerical examples optimal control, parameter identification shape problems PDEs are provided validate proposed theory illustrate ISMO significantly outperforms network
منابع مشابه
An Alternative Approach for PDE-Constrained Optimization via Genetic Algorithm
This paper deals with an alternative approach for solving the PDE-constrained optimization problems. For this purpose, the problem has been discretized with the help of finite difference method. Then the reduced problem has been solved by advanced real coded genetic algorithm with ranking selection, whole-arithmetic crossover and non uniform mutation. The proposed approach has been illustrated ...
متن کاملA Hybrid Optimization Algorithm for Learning Deep Models
Deep learning is one of the subsets of machine learning that is widely used in Artificial Intelligence (AI) field such as natural language processing and machine vision. The learning algorithms require optimization in multiple aspects. Generally, model-based inferences need to solve an optimized problem. In deep learning, the most important problem that can be solved by optimization is neural n...
متن کاملModel Order Reduction for PDE Constrained Optimization
The optimization and control of systems governed by partial differential equations (PDEs) usually requires numerous evaluations of the forward problem or the optimality system. Despite the fact that many recent efforts, many of which are reported in this book, have been made to limit or reduce the number of evaluations to 5–10, this cannot be achieved in all situations and even if this is possi...
متن کاملA Hybrid Optimization Algorithm for Learning Deep Models
Deep learning is one of the subsets of machine learning that is widely used in Artificial Intelligence (AI) field such as natural language processing and machine vision. The learning algorithms require optimization in multiple aspects. Generally, model-based inferences need to solve an optimized problem. In deep learning, the most important problem that can be solved by optimization is neural n...
متن کاملAlgorithms for PDE-Constrained Optimization
In this paper we review a number of algorithmic approaches for solving optimization problems with PDE constraints. Most of these methods were originally developed for finite dimensional problems. When applied to optimization problems with PDE constraints, new aspects become important. For instance, (discretized) PDE-constrained problems are inherently large-scale. Some aspects, like mesh indepe...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Computer Methods in Applied Mechanics and Engineering
سال: 2021
ISSN: ['0045-7825', '1879-2138']
DOI: https://doi.org/10.1016/j.cma.2020.113575