Table 3. Descriptor model.

Calibrated classifier performance in the training data (10-fold cross validated), the calculated probabilities are sorted into 10 equally sized bins. Additionally, the results from applying the calibrated descriptor model to the validation data are given.

bin	training data			validation data
	n benign	n malignant	cancer yield	n benign	n malignant	cancer yield
1	128	0	0	80	2	0.02439
2	128	0	0	95	0	0
3	128	0	0	191	5	0.02551
4	124	4	0.03125	102	2	0.019231
5	128	0	0	137	6	0.041958
6	126	2	0.015625	101	7	0.064815
7	119	9	0.070313	82	12	0.12766
8	110	18	0.140625	79	26	0.247619
9	94	34	0.265625	100	27	0.212598
10	53	71	0.572581	35	88	0.715447