BMC Medical Imaging

Table 2 Sensitivity analysis for alpha (accuracy)

From: A predicted-loss based active learning approach for robust cancer pathology image analysis in the workplace

Noisy dataset								Clean dataset
a	No threshold (LL)	0.5	0.8	1	Proposed	1.1	1.2	LL	Proposed
1	0.810 ± 0.012	0.812 ± 0.010	0.808 ± 0.014	0.815 ± 0.015	0.803 ± 0.007	0.811 ± 0.009	0.808 ± 0.017	0.805 ± 0.004	0.809 ± 0.004
2	0.794 ± 0.040	0.861 ± 0.006	0.817 ± 0.028	0.850 ± 0.020	0.856 ± 0.027	0.833 ± 0.025	0.839 ± 0.023	0.786 ± 0.124	0.846 ± 0.010
3	0.821 ± 0.035	0.808 ± 0.137	0.847 ± 0.030	0.868 ± 0.014	0.864 ± 0.032	0.842 ± 0.034	0.863 ± 0.015	0.855 ± 0.042	0.873 ± 0.013
4	0.815 ± 0.064	0.866 ± 0.011	0.876 ± 0.020	0.880 ± 0.014	0.881 ± 0.026	0.863 ± 0.017	0.860 ± 0.019	0.897 ± 0.012	0.892 ± 0.011
5	0.857 ± 0.044	0.675 ± 0.222	0.892 ± 0.010	0.881 ± 0.013	0.893 ± 0.018	0.892 ± 0.005	0.876 ± 0.029	0.827 ± 0.098	0.902 ± 0.019
6	0.859 ± 0.018	0.898 ± 0.010	0.879 ± 0.015	0.885 ± 0.028	0.899 ± 0.009	0.897 ± 0.009	0.887 ± 0.012	0.894 ± 0.027	0.892 ± 0.031
7	0.843 ± 0.048	0.870 ± 0.027	0.879 ± 0.035	0.905 ± 0.005	0.897 ± 0.021	0.893 ± 0.014	0.888 ± 0.011	0.903 ± 0.007	0.910 ± 0.009
8	0.897 ± 0.019	0.894 ± 0.025	0.906 ± 0.004	0.901 ± 0.014	0.908 ± 0.007	0.892 ± 0.022	0.873 ± 0.020	0.907 ± 0.010	0.912 ± 0.015
9	0.882 ± 0.045	0.882 ± 0.033	0.894 ± 0.021	0.904 ± 0.015	0.914 ± 0.007	0.878 ± 0.045	0.895 ± 0.010	0.882 ± 0.033	0.908 ± 0.019
10	0.863 ± 0.062	0.890 ± 0.019	0.898 ± 0.014	0.917 ± 0.006	0.917 ± 0.004	0.899 ± 0.016	0.888 ± 0.008	0.896 ± 0.014	0.924 ± 0.005

Table 2 shows the change in the accuracy of the model according to the change in the alpha value. We obtained the mean and standard deviation at each iteration after five trials for one alpha value. One trial includes 10 iterations and after 10 iterations, the next trial can be processed
LL Learning loss

Back to article page

ISSN: 1471-2342

Contact us

General enquiries: journalsubmissions@springernature.com