Data sovereignty compliant machine learning
The subject disclosure relates to systems for managing the deployment and updating of incremental machine learning models across multiple geographic sovereignties. In some aspects, systems of the subject technology are configured to perform operations including: receiving a first machine learning model via a first coordination agent, the first machine learning model based on a first training data set corresponding with a first sovereign region, sending the first machine learning model to a second coordination agent in a second sovereign region, wherein the second sovereign region is different from the first sovereign region, and receiving a second machine learning model from the second coordination agent, wherein the second machine learning model is based on updates to the first machine learning model using a second training data set corresponding with the second sovereign region. Methods and machine-readable media are also provided.
1. A coordination server configured to coordinate distribution of an incremental machine learning model, the coordination server comprising:
one or more processors;
a network interface coupled to the one or more processors; and
a memory coupled to the one or more processors, the memory comprising instructions stored therein, which when executed by the processors, cause the processors to perform operations comprising:
receiving a first machine learning model via a first coordination agent, the first machine learning model based on a first training data set corresponding with a first sovereign region;
sending the first machine learning model to a second coordination agent in a second sovereign region, wherein the second sovereign region is different from the first sovereign region; and
receiving a second machine learning model from the second coordination agent, wherein the second machine learning model is based on updates to the first machine learning model using a second training data set corresponding with the second sovereign region;
wherein the first sovereign region maintains export controls that prevent the first training set being sent to the coordination server and the second sovereign region;
wherein the second sovereign region maintains export controls that prevent the second training set being sent to the coordination server and the first sovereign region.
2. The coordination server of claim 1 , wherein the processors are further configured to perform operations comprising:
sending the second machine learning model to a third coordination agent located in a third sovereign region, wherein the third sovereign region is different from each of the first sovereign region and the second sovereign region; and
receiving a third machine learning model from the third coordination agent, wherein the third machine learning model is based on updates to the second machine learning model using a third training data set corresponding with the third sovereign region.
3. The coordination server of claim 1 , wherein the first coordination agent is located within the first sovereign region, and the second coordination agent is located within the second sovereign region.
4. The coordination server of claim 1 , wherein the first training data set is comprised of data for one or more users within the first sovereign region.
5. The coordination server of claim 1 , wherein the first machine learning model is an incremental machine learning model initially configured using a first training data set, and subsequently updated using a first new data set.
6. The coordination server of claim 1 , wherein the first machine learning model is configured to implement at least one of: a Naive Bayes classifier, a Bernoulli naive Bayes classifier, a Perceptron, a Stochastic Gradient Descent classifier, and a Passive Aggressive Classifier.
7. The coordination server of claim 1 , wherein the first machine learning model is configured to implement one or more of: a Stochastic Gradient Descent Regressor, and a Passive Aggressive Aggressor.
8. A computer-implemented method comprising:
receiving, at a coordination server, a first machine learning model via a first coordination agent, the first machine learning model based on a first training data set corresponding with a first sovereign region;
sending, by the coordination server, the first machine learning model to a second coordination agent in a second sovereign region, wherein the second sovereign region is different from the first sovereign region; and
receiving, by the coordination server, a second machine learning model from the second coordination agent, wherein the second machine learning model is based on updates to the first machine learning model using a second training data set corresponding with the second sovereign region;
wherein the first sovereign region maintains export controls that prevent the first training set being sent to the coordination server and the second sovereign region;
wherein the second sovereign region maintains export controls that prevent the second training set being sent to the coordination server and the first sovereign region.
9. The method of claim 8 , further comprising:
sending, by the coordination server, the second machine learning model to a third coordination agent located in a third sovereign region, wherein the third sovereign region is different from each of the first sovereign region and the second sovereign region; and
receiving a third machine learning model from the third coordination agent, wherein the third machine learning model is based on updates to the second machine learning model using a third training data set corresponding with the third sovereign region.
10. The method of claim 8 , wherein the first coordination agent is located within the first sovereign region, and the second coordination agent is located within the second sovereign region.
11. The method of claim 8 , wherein the first training data set is comprised of data for one or more users within the first sovereign region.
12. The method of claim 8 , wherein the first machine learning model is an incremental machine learning model initially configured using a first training data set, and subsequently updated using a first new data set.
13. The method of claim 8 , wherein the first machine learning model is configured to implement at least one of: a Naive Bayes classifier, a Bernoulli naive Bayes classifier, a Perceptron, a Stochastic Gradient Descent classifier, and a Passive Aggressive Classifier.
14. The method of claim 8 , wherein the first machine learning model is configured to implement one or more of: a Stochastic Gradient Descent Regressor, and a Passive Aggressive Aggressor.
15. A non-transitory computer-readable storage medium comprising instructions stored therein, which when executed by one or more processors, cause the processors to perform operations comprising:
receiving, at a coordination server, a first machine learning model via a first coordination agent, the first machine learning model based on a first training data set corresponding with a first sovereign region;
sending, by the coordination server, the first machine learning model to a second coordination agent in a second sovereign region, wherein the second sovereign region is different from the first sovereign region; and
receiving, by the coordination server, a second machine learning model from the second coordination agent, wherein the second machine learning model is based on updates to the first machine learning model using a second training data set corresponding with the second sovereign region;
wherein the first sovereign region maintains export controls that prevent the first training set being sent to the coordination server and the second sovereign region;
wherein the second sovereign region maintains export controls that prevent the second training set being sent to the coordination server and the first sovereign region.
16. The non-transitory computer-readable storage medium of claim 15 , further comprising:
sending, by the coordination server, the second machine learning model to a third coordination agent located in a third sovereign region, wherein the third sovereign region is different from each of the first sovereign region and the second sovereign region; and
receiving a third machine learning model from the third coordination agent, wherein the third machine learning model is based on updates to the second machine learning model using a third training data set corresponding with the third sovereign region.
17. The non-transitory computer-readable storage medium of claim 15 , wherein the first coordination agent is located within the first sovereign region, and the second coordination agent is located within the second sovereign region.
18. The non-transitory computer-readable storage medium of claim 15 , wherein the first training data set is comprised of data for one or more users within the first sovereign region.
19. The non-transitory computer-readable storage medium of claim 15 , wherein the first machine learning model is an incremental machine learning model initially configured using a first training data set, and subsequently updated using a first new data set.
20. The non-transitory computer-readable storage medium of claim 15 , wherein the first machine learning model is configured to implement at least one of: a Naive Bayes classifier, a Bernoulli naive Bayes classifier, a Perceptron, a Stochastic Gradient Descent classifier, and a Passive Aggressive Classifier.