ActiveLearningGaussianProcess

Description

The theory behind Gaussian Process (GP) is described in GaussianProcessTrainer. ActiveLearningGaussianProcess is slightly similar to the GaussianProcessTrainer class in that it trains a GP model. However, a key feature of ActiveLearningGaussianProcess is that it permits re-training the GP model on-the-fly during the active learning process. This means that the input the inputs and output data set sizes will be dynamic and re-training the GP can be performed several times as dictated by the learning (or acquisition) function during active learning.

Just like the GaussianProcessTrainer class, the GP model during active learning can be trained using either of the following options:

  • PETSc/TAO

    Relies on the TAO optimization library from PETSc. Several optimization algorithms are available from this library. Note that these algorithms perform deterministic optimization.

  • Adaptive moment estimation (Adam)

    Relies on the pseudocode provided in Kingma and Ba (2014). Adam permits stochastic optimization, wherein, a batch of the training data can be randomly chosen at each iteration.

Interaction between ActiveLearningMonteCarloSampler, ActiveLearningGaussianProcess, and ActiveLearningGPDecision

Schematic of active learning in Monte Carlo simulations with parallel computing. The interaction between the three objects, `ActiveLearningMonteCarloSampler`, `ActiveLearningGaussianProcess`, and `ActiveLearningGPDecision`, is presented.

Figure 1: Schematic of active learning in Monte Carlo simulations with parallel computing. The interaction between the three objects, ActiveLearningMonteCarloSampler, ActiveLearningGaussianProcess, and ActiveLearningGPDecision, is presented.

Usage of active learning

Please refer to ActiveLearningGPDecision on a detailed description on using active learning.

warningwarning

The detailed documentation of this object is only available when Moose is compiled with Libtorch. For instructions on how to compile Moose with Libtorch, visit the general installation webpage or click here.

References

  1. D. P. Kingma and J. Ba. Adam: a method for stochastic optimization. arXiv:1412.6980 [cs.LG], 2014. URL: https://doi.org/10.48550/arXiv.1412.6980.[Export]