TB Hyperparameter optimization for GaAs

This example demonstrates hyperparameter optimization of a TB model for GaAs using a hybrid orbital basis. While all numerical input parameters can in principle be optimized, we focus here on the distance-dependence parameter alpha. All input files can be found here.

To run a hyperparameter optimization, both an Optimizer and a HyperOpt block must be present in the config file. At minimum, the parameters to be optimized and their respective search windows (lowerbounds and upperbounds) need to be specified. The search interval can further be refined using the stepsizes argument.

Three hyperparameter optimization methods are available: Random Search (mode=random), Grid Search (mode=grid), and the Tree-structured Parzen Estimator (mode=tpe).

As additional input files, two data sets (training and validation) as well as a structure file are required.

By default, this workflow reports an update only after each completed hyperparameter optimization step, showing the final validation loss and the current optimum for comparison. An estimate of the remaining runtime is also provided, based on the average iteration time.