Neural network architecture for attention based efficient model adaptation

1. A neural network system comprising:

processing hardware; and

a memory storing instructions which, when executed by the processing hardware, cause the processing hardware to perform operations comprising:

accessing an input vector, the input vector comprising a numeric representation of an input to a neural network;

providing the input vector to the neural network comprising a plurality of ordered layers, wherein each layer in at least a subset of the plurality of ordered layers is coupled with an adaptation module, wherein the adaptation module receives a same input value as a coupled layer for the adaptation module, and wherein an output value of the adaptation module is pointwise multiplied with an output value of the coupled layer to generate a next layer input value, wherein the adaptation module is trained using different training data from the ordered layers of the neural network; and

generating an output of the neural network based on an output of a last one of the plurality of ordered layers in the neural network.

2. The system of claim 1 , wherein the adaptation module is lightweight relative to the coupled layer.

3. The system of claim 1 , wherein the adaptation module is trained separately from the plurality of ordered layers of the neural network.

4. The system of claim 3 , wherein the plurality of ordered layers of the neural network are trained to solve a first problem, and wherein the adaptation module, coupled with the neural network, is trained to solve a second problem different from the first problem.

5. The system of claim 4 , wherein the adaptation module guides the neural network to focus on input regions that are relevant to the second problem.

6. The system of claim 4 , wherein the first problem is image recognition.

7. The system of claim 4 , \herein the generated output comprises a solution to the second problem.

8. The system of claim 1 , wherein the at least the subset of the plurality of ordered layers comprises a last m layers of the plurality of ordered layers, wherein m is a positive integer.

9. A non-transitory machine-readable medium storing instructions which, when executed by one or more computing machines, cause the one or more computing machines to perform operations comprising:

accessing an input vector, the input vector comprising a numeric representation of an input to a neural network;

providing the input vector to the neural network comprising a plurality of ordered layers, wherein each layer in at least a subset of the plurality of ordered layers is coupled with an adaptation module, wherein the adaptation module receives a same input value as a coupled layer for the adaptation module, and wherein an output value of the adaptation module is pointwise multiplied with an output value of the coupled layer to generate a next layer input value, wherein the adaptation module is trained using different training data from the ordered layers of the neural network; and

generating an output of the neural network based on an output of a last one of the plurality of ordered layers in the neural network.

10. The machine-readable medium of claim 9 , wherein the adaptation module is lightweight relative to the coupled layer.

11. The machine-readable medium of claim 9 , wherein the adaptation module is trained separately from the plurality of ordered layers of the neural network.

12. The machine-readable medium of claim 11 , wherein the plurality of ordered layers of the neural network are trained to solve a first problem, and wherein the adaptation module, coupled with the neural network, is trained to solve a second problem different from the first problem.

13. The machine-readable medium of claim 12 , wherein the adaptation module guides the neural network to focus on input regions that are relevant to the second problem.

14. The machine-readable medium of claim 12 , wherein the first problem is image recognition.

15. The machine-readable medium of claim 12 , wherein the generated output comprises a solution to the second problem.

16. The machine-readable medium of claim 9 , wherein the at least the subset of the plurality of ordered layers comprises a last m layers of the plurality of ordered layers, wherein m is a positive integer.

17. A neural network method comprising:

accessing, at one or more computing machines, an input vector, the input vector comprising a numeric representation of an input to a neural network stored at the one or more computing machines;

providing the input vector to the neural network comprising a plurality of ordered layers, wherein each layer in at least a subset of the plurality of ordered layers is coupled with an adaptation module, wherein the adaptation module receives a same input value as a coupled layer for the adaptation module, and wherein an output value of the adaptation module is pointwise multiplied with an output value of the coupled layer to generate a next layer input value, wherein the adaptation module is trained using different training data from the ordered layers of the neural network; and

generating, at the one or more computing machines, an output of the neural network based on an output of a last one of the plurality of ordered layers in the neural network.

18. The method of claim 17 , wherein the adaptation module is lightweight relative to the coupled layer.

19. The method of claim 17 ; wherein the adaptation module is trained separately from the plurality of ordered layers of the neural network.

20. The method of claim 19 , wherein the plurality of ordered layers of the neural network are trained to solve a first problem, and wherein the adaptation module, coupled with the neural network, is trained to solve a second problem different from the first problem.