Table 3.
Uncertainty - error decrease: Performance of various ensemble models using the proposed uncertainty approach
Metric | % kept epochs | ||||
---|---|---|---|---|---|
90 | 75 | 50 | 25 | 10 | |
Avg. Err. Decrease | 3.5 | 5.4 | 6.6 | 5.5 | 2.9 |
Avg. Max Err. Decrease | 5.5 | 9.5 | 13.6 | 16.5 | 16.3 |
Max Err. Decrease | 9.9 | 21.4 | 32.0 | 31.9 | 24.0 |
The error decrease is calculated from the improvement in accuracy and is presented as the “percentage of error decrease”. The “Avg. Error Decrease” signifies the average error reduction across all ensemble models and iterations (folds). “Average Maximum” denotes the mean value of the largest error reductions observed across all ensembles. The “Maximum Error Decrease” refers to the highest error reduction observed across all ensemble models. This visualization allows for the quantification of the impact of the uncertainty method on performance