Table 3.

Weighted F1-score on the testing set for the linear SVM classifier trained on reference sets generated from each round of automatic expansion.

Datasets	Rounds	1	2	3	4	5	6	7	8	9	10
Ovarian T cells	Ref. Size	6	12	18	24	30	36	42	48	54	60
	Weighted F1-score	0.762 (0.007)	0.889(0.015)	0.891(0.016)	0.881 (0.02)	0.884 (0.008)	0.866 (0.016)	0.877 (0.012)	0.879 (0.019)	0.877 (0.019)	0.87 (0.02)
LUSC T cells	Ref. Size	4	8	12	16	20	24	28	32	36	40
	Weighted F1-score	0.787 (0.064)	0.904 (0.033)	0.907 (0.017)	0.908 (0.015)	0.922(0.011)	0.918 (0.027)	0.921(0.021)	0.918 (0.02)	0.918 (0.014)	0.912 (0.004)
DCIS FOXP3	Ref. Size	2	4	6	8	10	12	14	16	18	20
	Weighted F1-score	0.775 (0.008)	0.984(0.003)	0.977 (0.008)	0.975 (0.001)	0.974 (0.002)	0.962 (0.011)	0.968 (0.015)	0.965 (0.017)	0.967 (0.016)	0.975 (0.008)
Myeloma	Ref. Size	3	6	9	12	15	18	21	24	27	30
	Weighted F1-score	0.672 (0.2590)	0.742 (0.131)	0.743 (0.156)	0.839 (0.091)	0.919 (0.028)	0.912 (0.025)	0.918 (0.026)	0.93 (0.012)	0.936 (0.006)	0.948(0.001)
IMC CD4-CD8	Ref. Size	2	4	6	8	10	12	14	16	18	20
	Weighted F1-score	0.584 (0.147)	0.801 (0.11)	0.853 (0.005)	0.857 (0.024)	0.863 (0.021)	0.863 (0.015)	0.906 (0.029)	0.899 (0.036)	0.912 (0.012)	0.917(0.015)

The average of 3 repeats with different initial reference sets is shown, with standard deviations shown inside the parentheses.

Bold values are within 0.003 lower than the best.