Cascaded neural networks using test ouput from the first neural network to train the second neural network

1. A method, implemented by a computer, comprising:

training a first neural network using a first training dataset;

inputting each test data of a first test dataset to the first neural network;

calculating output data of the first neural network for each test data of the first test dataset;

composing a second training dataset of training data from the first test dataset that causes the first neural network to output data within a first range; and

training a second neural network using the second training dataset.

2. The method of claim 1 , wherein the first test dataset is the same as the first training dataset.

3. The method of claim 1 , wherein the first test dataset is different from the first training dataset.

4. The method of claim 1 , wherein

the training of the first network comprises training each first neural network of a plurality of first neural networks by using a respective first training dataset of a plurality of the first training datasets, and

the training of the second neural network comprises training each second network of a plurality of the second neural networks by using a respective second training dataset of a plurality of the second training datasets.

5. The method of claim 4 , further comprising:

obtaining an initial training dataset;

dividing the initial training dataset into a plurality of groups; and

generating the plurality of first training datasets by allocating at least one group to each first training dataset, wherein different combinations of groups among the plurality of groups are allocated to each first training dataset.

6. The method of claim 1 , wherein the first range is a range on or above a threshold.

7. The method of claim 6 , further comprising:

inputting each test data of a second test dataset to the first neural network; and

determining the threshold based on an average and a standard deviation of output data of the first neural network for the second test dataset.

8. The method of claim 1 , wherein the first range is a range below a threshold.

9. The method of claim 1 , wherein each training data of the first training dataset and each test data of the first test dataset includes an image and a classification of the image.

10. The method of claim 9 , wherein the classification of the image includes a first attribute and a second attribute, and

wherein the number of training data having the second attribute is at least 4 times larger than the number of training data having the first attribute in the first training dataset.

11. The method of claim 10 , wherein the images of the first training dataset include a plurality of cross-sectional images of a body tissue.

12. The method of claim 11 , wherein the images of the first test dataset are obtained by slicing a body tissue of the images of the first training dataset at a plurality of cross-sections that are different than the cross-sections of the images of the first training dataset.

13. The method of claim 1 , further comprising:

inputting each target data of a first target dataset to the first neural network; and

calculating output data of the first neural network for each target data of the first target dataset.

14. The method of claim 13 , further comprising:

composing a second target dataset of target data from the first target dataset that causes the first neural network to output data within a second range; and

calculating output data of the second neural network for each target data of the second target dataset.

15. The method of claim 14 , further comprising:

adopting highest output data of the first neural network for an image of the first target dataset;

wherein the first target dataset includes a plurality of images of a body tissue, the plurality of images having at least one of a different coverage, different angle, and different magnification ratio.

16. A system, comprising:

a processor; and

one or more computer readable mediums collectively including instructions that, when executed by the processor, cause the processor to:

train a first neural network using a first training dataset,

input each test data of a first test dataset to the first neural network,

calculate output data of the first neural network for each test data of the first test dataset,

compose a second training dataset of training data from the first test dataset that causes the first neural network to output data within a first range, and

train a second neural network using the second training dataset.

17. The system of claim 16 , wherein the first test dataset is the same as the first training dataset.

18. The system of claim 16 , wherein the first test dataset is different from the first training dataset.

19. A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform operations comprising:

training a first neural network using a first training dataset;

inputting each test data of a first test dataset to the first neural network;

calculating output data of the first neural network for each test data of the first test dataset;

composing a second training dataset of training data from the first test dataset that causes the first neural network to output data within a first range; and

training a second neural network using the second training dataset.

20. The apparatus of claim 16 , wherein the first test dataset is the same as the first training dataset.