Figure 17. (left) Model prediction errors for NSTX validation shots using the latest available online learning model compared with the model trained on all NSTX training shots. (right) Distribution of results for both cases (shaded regions). In these plots, white dots indicate medians, thick lines indicate the interval between first and third quartiles, and thin lines represent the interval between lower and upper adjacent values. Median prediction error increases when predicting based on available data and the worst predictions, while rare, are significantly worse than for the model trained on all data.