Changeover from one neural network to another neural network

1. A computerized method for changing a system over from a first deep convolutional neural network (DCNN) to a second DCNN, the first DCNN trained to output a value for a first metric for a document input to the first DCNN and the second DCNN trained to output a transformation of a first transformation type for a document input to the first DCNN, the method comprising:

for each of one or more input documents:

determining a first value for the first metric for a first transformation of the input document by passing the first transformation to the first DCNN;

obtaining a second transformation of the input document by passing the input document to the second DCNN;

passing the second transformation of the input document to the first DCNN, obtaining a second value for the first metric for the second transformation of the input document, the first and second transformations being of the first transformation type;

determining, for the input document, a difference between the first value and the second value;

determining whether to change the system over from the first DCNN to the second DCNN based on the difference between the first value and the second value for each of the one or more input documents; and

in response to a determination that the system should be changed over from the first DCNN to the second DCNN, causing, for transformations of the first transformation type, new input documents to the system to be passed to the second DCNN in lieu of being passed to the first DCNN.

2. The method of claim 1 , wherein the determining whether to change the system over includes adding a distance factor to the difference, the distance factor set based on savings of computational power by virtue of using the second DCNN in lieu of the first DCNN.

3. The method of claim 1 , wherein the first metric is a professionalism score.

4. The method of claim 1 , wherein the determining whether to change the system over includes passing two versions of the same input document to the second DCNN and determining whether the second DCNN outputs the same second transformation for each of the two versions.

5. The method of claim 4 , wherein the two versions of the same input document are partially overlapping portions of a single image.

6. The method of claim 1 , wherein the first transformation is obtained by feeding a plurality of possible transformations of the first input document to the first DCNN and, for each of the plurality of possible transformations, within the first DCNN:

passing the transformation to the convolutional layer, generating output;

passing the output from the convolutional layer to the nonlinearity layer, generating output;

passing the output from the nonlinearity layer to the pooling layer, generating output; and

passing output from the nonlinearity layer to the classification layer, generating a professionalism score for the transformation.

7. The method of claim 1 , wherein the transformation type is a cropping.

8. A system comprising:

a computer readable medium having instructions stored there on, which, when executed by a processor, cause the system to perform a method for changing a system over from a first deep convolutional neural network (DCNN) to a second DCNN, the first DCNN trained to output a value for a first metric for a document input to the first DCNN and the second DCNN trained to output a transformation of a first transformation type for a document input to the first DCNN, by:

for each of one or more input documents:

determining a first value for the first metric for a first transformation of the input document by passing the first transformation to the first DCNN;

obtaining a second transformation of the input document by passing the input document to the second DCNN;

passing the second transformation of the input document to the first DCNN, obtaining a second value for the first metric for the second transformation of the input document, the first and second transformations being of the first transformation type;

determining, for the input document, a difference between the first value and the second value;

determining whether to change the system over from the first DCNN to the second DCNN based on the difference between the first value and the second value for each of the one or more input documents; and

in response to a determination that the system should be changed over from the first DCNN to the second DCNN, causing, for transformations of the first transformation type, new input documents to the system to be passed to the second DCNN in lieu of being passed to the first DCNN.

9. The system of claim 8 , wherein the determining whether to change the system over includes adding a distance factor to the difference, the distance factor set based on savings of computational power by virtue of using the second DCNN in lieu of the first DCNN.

10. The system of claim 8 , wherein the first metric is a professionalism score.

11. The system of claim 8 , wherein the determining whether to change the system over includes passing two versions of the same input document to the second DCNN and determining whether the second DCNN outputs the same second transformation for each of the two versions.

12. The system of claim 11 , wherein the two versions of the same input document are partially overlapping portions of a single image.

13. The system of claim 8 , wherein the first transformation is obtained by feeding a plurality of possible transformations of the first input document to the first DCNN and, for each of the plurality of possible transformations, within the first DCNN:

passing the transformation to the convolutional layer, generating output;

passing the output from the convolutional layer to the nonlinearity layer, generating output;

passing the output from the nonlinearity layer to the pooling layer, generating output; and

passing output from the nonlinearity layer to the classification layer, generating a professionalism score for the transformation.

14. The system of claim 8 , herein the transformation type is a cropping.

15. A non-transitory machine-readable storage medium comprising instructions, which when implemented by one or more machines, cause the one or more machines to perform operations for changing a system over from a first deep convolutional neural network (DCNN) to a second DCNN, the first DCNN trained to output a value for a first metric for a document input to the first DCNN and the second DCNN trained to output a transformation of a first transformation type for a document input to the first DCNN, the operations comprising:

for each of one or more input documents:

determining a first value for the first metric for a first transformation of the input document by passing the first transformation to the first DCNN;

obtaining a second transformation of the input document by passing the input document to the second DCNN;

passing the second transformation of the input document to the first DCNN, obtaining a second value for the first metric for the second transformation of the input document, the first and second transformations being of the first transformation type;

determining, for the input document, a difference between the first value and the second value;

determining whether to change the system over from the first DCNN to the second DCNN based on the difference between the first value and the second value for each of the one or more input documents; and

in response to a determination that the system should be changed over from the first DCNN to the second DCNN, causing, for transformations of the first transformation type, new input documents to the system to be passed to the second DCNN in lieu of being passed to the first DCNN.

16. The non-transitory machine-readable storage medium of claim 15 , wherein the determining whether to change the system over includes adding a distance factor to the difference, the distance factor set based on savings of computational power by virtue of using the second DCNN in lieu of the first DCNN.

17. The non-transitory machine-readable storage medium of claim 15 , wherein the first metric is a professionalism score.

18. The non-transitory machine-readable storage medium of claim 15 , wherein the determining whether to change the system over includes passing two versions of the same input document to the second DCNN and determining whether the second DCNN outputs the same second transformation for each of the two versions.

19. The non-transitory machine-readable storage medium of claim 18 , wherein the two versions of the same input document are partially overlapping portions of a single image.

20. The non-transitory machine-readable storage medium of claim 15 , wherein the first transformation is obtained by feeding a plurality of possible transformations of the first input document to the first DCNN and, for each of the plurality of possible transformations, within the first DCNN:

passing the transformation to the convolutional layer, generating output;

passing the output from the convolutional layer to the nonlinearity layer, generating output;

passing the output from the nonlinearity layer to the pooling layer, generating output; and

passing output from the nonlinearity layer to the classification layer, generating a professionalism score for the transformation.