Parametric Optimization using Neural Networks as Surrogates
Start using Neural Networks not just for prediction, but for optimization. In many engineering and scientific problems, we often deal with “black box” functions where the relationship between inputs and outputs is complex or expensive to compute (e.g., fluid dynamics simulations, finite element analysis). We want to find the input parameters that minimize (or maximize) some objective, but we can’t easily calculate gradients to guide us. This is where Neural Networks as Surrogate Models come in. ...