Robust sparse linear regression for tokamak plasma boundary estimation using variational Bayes

Abstract

Abstract. Precise control of the shape of plasma in a tokamak requires reliable reconstruction of the plasma boundary. The problem of boundary estimation can be reduced to a simple linear regression with a potentially infinite amount of regressors. This regression problem poses some difficulties for classical methods. The selection of regressors significantly influences the reconstructed boundary. Also, the underlying model may not be valid during certain phases of the plasma discharge. Formal model structure estimation technique based on the automatic relevance principle yields a version of sparse least squares estimator. In this contribution, we extend the previous method by relaxing the assumption of Gaussian noise and using Student's t-distribution instead. Such a model is less sensitive to potential outliers in the measurement. We show on simulations and real data that the proposed modification improves estimation of the plasma boundary in some stages of a plasma discharge. Performance of the resulting algorithm is evaluated with respect to a more detailed and computationally costly model which is considered to be the "ground truth". The results are also compared to those of Lasso and Tikhonov regularization techniques.

Neural network surrogate of QuaLiKiz using JET experimental data to populate training space

“…With the development of increasingly powerful high performance computing resources, machine learning (ML) techniques are becoming practical tools in model construction and large data exploitation. These tools open new exploratory options to address current issues in fusion plasma analysis and operation [1][2][3][4][5]. Specifically, the usage of neural networks (NNs) [6] as fast surrogate models has already been applied to plasma microturbulent transport calculations [7,8].…”

“…The reference time, t = 0, in the JET data system is when the magnetic coils start ramping up, instead of the usual time of plasma breakdown. These two events are typically 40 s apart at JET 4. The particle transport options are only applicable when using the QLKNN model.…”

Efficient training sets for surrogate models of tokamak turbulence with Active Deep Ensembles

“…Active Learning is a sequential sampling strategy that queries an expensive black box function (in our case a gyrokinetic model) by means of an acquisition function that identifies regions where additional data would improve the NN performance. Contrary to Bayesian Optimisation approaches, which aim to perform sequential optimisation with only a few function evaluations (see for example [17,[27][28][29] for applications relevant to Fusion), Active Learning enables learning of the function to be approximated over the entire parameter space.…”