Figure 5.

(a) Completion ratio of evaluation runs. Runs involving machine and user co-adaptation, regardless of algorithm or forgetting factor, had, on average, higher target hit rates compared to initial evaluations with the batch-trained model (Baseline). (b) Averaged dot products between normalised row vectors of weights from batch-trained models and their online-adapted versions. A value of 1 indicates no change to the sensitivities of the mapping while lower values represent larger changes to the model during machine adaptation.