Hybrid quantized decision model framework

1. A system, comprising:

at least one computing device; and

program instructions stored in at least one memory of the at least one computing device, wherein the instructions, when executed by at least one processor, cause the at least one computing device to at least:

identify, by a cloud computing environment executing a cloud decision manager, a cloud layer decision model that generates cloud layer predictions based on device layer data for at least one device;

generate, by the cloud computing environment, a quantized decision model that is a quantized version of the cloud layer decision model, the quantized decision model comprising a decreased computational requirement relative to the cloud layer decision model;

identify, by the cloud computing environment, a logical group of at least one edge device that collects the device layer data from the at least one device; and

transmit, by the cloud computing environment, the quantized decision model to the at least one edge device along with a task profile that specifies a round-trip network latency threshold that causes the at least one edge device to identify a current edge-cloud round-trip network latency and use the quantized decision model in an instance in which the current edge-cloud round-trip network latency is greater than the round-trip network latency threshold.

2. The system of claim 1 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

generate, by the cloud computing environment, the task profile comprising a unique identifier of the quantized decision model, the logical group of the at least one edge device, and an edge layer decision threshold that indicates a prediction confidence for an edge device to utilize the quantized decision model without consulting the cloud decision manager.

3. The system of claim 1 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

receive, by the cloud computing environment, a cloud layer prediction request from an edge device of the at least one edge device, the cloud layer prediction request specifying: the device layer data, and at least one of: the cloud layer decision model, or the quantized decision model; and

transmit, by the cloud computing environment, a cloud layer prediction generated based on the cloud layer decision model.

4. The system of claim 3 , wherein the cloud layer prediction request further specifies an identification of the edge device.

5. The system of claim 3 , wherein the cloud layer prediction further specifies an action.

6. The system of claim 3 , wherein the cloud layer prediction is transmitted as a verification of an edge layer prediction generated by the edge device using the quantized decision model.

7. The system of claim 3 , wherein the cloud layer prediction is transmitted as a correction of an edge layer prediction generated by the edge device using the quantized decision model.

8. A method, comprising:

identifying, by a cloud computing environment executing a cloud decision manager, a cloud layer decision model that generates cloud layer predictions based on device layer data for at least one device;

generating, by the cloud computing environment, a quantized decision model that is a quantized version of the cloud layer decision model, the quantized decision model comprising a decreased computational requirement relative to the cloud layer decision model;

identifying, by the cloud computing environment, a logical group of at least one edge device that collects the device layer data from the at least one device; and

transmitting, by the cloud computing environment, the quantized decision model to the at least one edge device along with a task profile that specifies a round-trip network latency threshold that causes the at least one edge device to identify a current edge-cloud round-trip network latency and use the quantized decision model in an instance in which the current edge-cloud round-trip network latency is greater than the round-trip network latency threshold.

9. The method of claim 8 , further comprising:

generating, by the cloud computing environment, the task profile comprising a unique identifier of the quantized decision model, the logical group of the at least one edge device, and an edge layer decision threshold that indicates a prediction confidence for an edge device to utilize the quantized decision model without consulting the cloud decision manager.

10. The method of claim 8 , further comprising:

receiving, by the cloud computing environment, a cloud layer prediction request from an edge device of the at least one edge device, the cloud layer prediction request specifying: the device layer data, and at least one of: the cloud layer decision model, or the quantized decision model; and

transmitting, by the cloud computing environment, a cloud layer prediction generated based on the cloud layer decision model.

11. The method of claim 10 , wherein the cloud layer prediction request further specifies an identification of the edge device.

12. The method of claim 10 , wherein the cloud layer prediction further specifies an action.

13. The method of claim 10 , wherein the cloud layer prediction is transmitted as a verification of an edge layer prediction generated by the edge device using the quantized decision model.

14. The method of claim 10 , wherein the cloud layer prediction is transmitted as a correction of an edge layer prediction generated by the edge device using the quantized decision model.

15. A non-transitory computer-readable medium comprising program instructions that when executed by at least one processor, cause at least one computing device to at least:

identify, by a cloud computing environment executing a cloud decision manager, a cloud layer decision model that generates cloud layer predictions based on device layer data for at least one device;

generate, by the cloud computing environment, a quantized decision model that is a quantized version of the cloud layer decision model, the quantized decision model comprising a decreased computational requirement relative to the cloud layer decision model;

identify, by the cloud computing environment, a logical group of at least one edge device that collects the device layer data from the at least one device; and

transmit, by the cloud computing environment, the quantized decision model to the at least one edge device along with a task profile that specifies a round-trip network latency threshold that causes the at least one edge device to identify a current edge-cloud round-trip network latency and use the quantized decision model in an instance in which the current edge-cloud round-trip network latency is greater than the round-trip network latency threshold.

16. The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

generate, by the cloud computing environment, the task profile comprising a unique identifier of the quantized decision model, the logical group of the at least one edge device, and an edge layer decision threshold that indicates a prediction confidence for an edge device to utilize the quantized decision model without consulting the cloud decision manager.

17. The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

receive, by the cloud computing environment, a cloud layer prediction request from an edge device of the at least one edge device, the cloud layer prediction request specifying: the device layer data, and at least one of: the cloud layer decision model, or the quantized decision model; and

transmit, by the cloud computing environment, a cloud layer prediction generated based on the cloud layer decision model.

18. The non-transitory computer-readable medium of claim 17 , wherein the cloud layer prediction further specifies an action.

19. The non-transitory computer-readable medium of claim 17 , wherein the cloud layer prediction is transmitted as a verification of an edge layer prediction generated by the edge device using the quantized decision model.

20. The non-transitory computer-readable medium of claim 17 , wherein the cloud layer prediction is transmitted as a correction or a verification of an edge layer prediction generated by the edge device using the quantized decision model.