Systems and methods for enhancing a distributed medical network

1. A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:

inputting training data corresponding to each local computer into a corresponding machine learning model;

generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data;

generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss;

generating a global parameter update based at least in part on the plurality of local parameter gradients;

updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update;

delivering the plurality of local parameter gradients from the plurality of local computers to a central computer of the distributed medical network; and

delivering the global parameter update from the central computer to the plurality of local computers;

wherein the generating a global parameter update includes generating the global parameter update at the central computer.

2. The computer-implemented method of claim 1 , wherein the local parameter gradient is one of a local weight gradient and a local bias gradient.

3. The computer-implemented method of claim 1 , wherein the generating a local parameter gradient includes determining a non-zero parameter gradient.

4. The computer-implemented method of claim 1 , wherein the generating the global parameter update includes averaging the plurality of local parameter gradients.

5. The computer-implemented method of claim 1 , further comprising:

before delivering the plurality of local parameter gradients, compressing and encrypting the plurality of local parameter gradients at the plurality of local computers; and

after delivering the plurality of local parameter gradients, decrypting and decompressing the compressed plurality of local parameter gradients at the central computer.

6. The computer-implemented method of claim 1 , further includes:

implementing a time out period on the central computer such that the central computer is configured to reject receipt of local parameter gradient after the time out period.

7. A system for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the system comprising:

a data inputting module implemented by one or more processors and configured to input training data corresponding to each local computer into a corresponding machine learning model;

a loss generating module implemented by the one or more processors and configured to generate a plurality of local losses by at least generating a local loss for each machine learning model based at least in part on the corresponding training data;

a gradient generating module implemented by the one or more processors and configured to generate a plurality of local parameter gradients by at least generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss;

an update generating module implemented by the one or more processors and configured to generate a global parameter update based at least in part on the plurality of local parameter gradients;

a model updating module implemented by the one or more processors and configured to update each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update; and

a delivering module implemented by the one or more processors and configured to:

deliver the plurality of local parameter gradients from the plurality of local computers to a central computer of the distributed medical network; and

deliver the global parameter update from the central computer to the plurality of local computers;

wherein the update generating module is further configured to generate the global parameter update at the central computer.

8. A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:

inputting training data corresponding to each local computer into a corresponding machine learning model;

generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data;

generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss;

generating a global parameter update based at least in part on the plurality of local parameter gradients;

updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update; and

delivering the plurality of local parameter gradients between the plurality of local computers;

wherein the generating a global parameter update includes generating the global parameter update at each local computer of the plurality of local computers.

9. The computer-implemented method of claim 8 , further comprising:

implementing a time out period on each local computer such that each local computer is configured to reject receipt of local parameter gradient after the time out period.

10. The computer-implemented method of claim 8 , further comprising:

before delivering the plurality of local parameter gradients, compressing and encrypting the plurality of local parameter gradients at the plurality of local computers; and

after delivering the plurality of local parameter gradients, decrypting and decompressing the compressed plurality of local parameter gradients at the plurality of local computers.

11. A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:

inputting training data corresponding to each local computer into a corresponding machine learning model;

generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data;

generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss;

generating a global parameter update based at least in part on the plurality of local parameter gradients;

updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update;

generating a plurality of local parameter sets including generating a local parameter set for each machine learning model based at least in part on a corresponding local parameter gradient of the plurality of local parameter gradients;

wherein the generating a global parameter update based at least in part on the plurality of local parameter gradients includes generating the global parameter update based at least in part on the plurality of local parameter sets;

delivering the plurality of local parameter sets from the plurality of local computers to a central computer of the distributed medical network; and

delivering the global parameter update from the central computer to the plurality of local computers;

wherein the generating a global parameter update includes generating the global parameter update at the central computer.

12. The computer-implemented method of claim 11 , wherein the local parameter set includes one of a local set of weights and a local set of biases.

13. The computer-implemented method of claim 11 , wherein the generating the global parameter update includes averaging the plurality of local parameter sets.

14. The computer-implemented method of claim 11 , further comprising:

before delivering the plurality of local parameter sets, compressing and encrypting the plurality of local parameter sets at the plurality of local computers; and

after delivering the plurality of local parameter sets, decrypting and decompressing the compressed plurality of local parameter sets at the central computer.

15. The computer-implemented method of claim 11 , further includes:

implementing a time out period on the central computer such that the central computer is configured to reject receipt of local parameter set after the time out period.

16. A non-transitory computer-readable medium with instructions stored thereon, that when executed by a processor, causes the processor to perform one or more processes including:

inputting training data corresponding to each local computer of a plurality of local computers into a corresponding machine learning model, each local computer of the plurality of local computers hosting a machine learning model;

generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data;

generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss;

generating a global parameter update based at least in part on the plurality of local parameter gradients; and

updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update; and

delivering the plurality of local parameter gradients between the plurality of local computers;

wherein the generating a global parameter update includes generating the global parameter update at each local computer of the plurality of local computers.

17. A computer-implemented method for enhancing a distributed medical network including a plurality of local computers, each local computer of the plurality of local computers hosting a machine learning model configured with an active parameter set, the method comprising:

inputting training data corresponding to each local computer into a corresponding machine learning model;

generating a plurality of local losses including generating a local loss for each machine learning model based at least in part on the corresponding training data;

generating a plurality of local parameter gradients including generating a local parameter gradient for each machine learning model based at least in part on the corresponding local loss;

generating a global parameter update based at least in part on the plurality of local parameter gradients;

updating each machine learning model hosted at each local computer of the plurality of local computers by at least updating their corresponding active parameter set based at least in part on the global parameter update;

generating a plurality of local parameter sets including generating a local parameter set for each machine learning model based at least in part on a corresponding local parameter gradient of the plurality of local parameter gradients;

wherein the generating a global parameter update based at least in part on the plurality of local parameter gradients includes generating the global parameter update based at least in part on the plurality of local parameter sets;

delivering the plurality of local parameter sets between the plurality of local computers;

wherein the generating a global parameter update includes generating the global parameter update at each local computer of the plurality of local computers.

18. The computer-implemented method of claim 17 , further comprising:

before delivering the plurality of local parameter sets, compressing and encrypting the plurality of local parameter sets at the plurality of local computers; and

after delivering the plurality of local parameter sets, decrypting and decompressing the compressed plurality of local parameter sets at the plurality of local computers.