We have repeated the above weight updating procedure until the threshold in terms of the required mean square error is achieved, where the threshold mean square error is the tolerable prediction error.
Surprisingly, we find that higher Uniform Threshold values increase how much link-prediction error is tolerable in planning complete cascades.
We say that a rate of link-prediction error is tolerable if (V') remains competitive with seeding based on perfectly accurate link information, and most realizations of (G') yield a (V') with performance that exceeds random seeding.
This is important because if a particular variable had very little influence on the predictive model, then larger imputation errors are tolerable, resulting in negligible loss of prediction accuracy.
Tolerable, however, is a relative notion.
As shown in Fig. 1B, the coincidence of prediction and measurement was satisfactory, with a tolerable systematic scatter of experimental values around the model prediction (for clarity, only one of the two experiments used for the calculations is plotted).
This equates to a high level of false alarms, which may be tolerable if the cost of confirming a prediction (e.g., secondary screening) is acceptably low.
The deviation of the model from the experimental data was estimated to be about 18% (squared deviations) or 15% (absolute deviations), which indicates a tolerable systematic scatter of experimental values around the model prediction.
The predictions D and T for SIFT mean Deleterious and Tolerable respectively.
This model predictions have been validated by proper tests, and based on this experimental evidence a criterion to accept tolerable short cracks has been proposed.
We expect that a good classifier should give especially precise predictions at the top levels of the KEGG hierarchy, while at the bottom levels misclassifications are more tolerable.
