Federated artificial intelligence with cumulative learning in a computer network
In one embodiment, a first deep fusion reasoning engine (DFRE) agent in a network receives first sensor data from a first set of one or more sensors in the network. The first DFRE agent translates the first sensor data into symbolic data. The first DFRE agent applies, using a symbolic knowledge base maintained by the first DFRE agent, symbolic reasoning to the symbolic data to make an inference regarding the first sensor data. The first DFRE agent updates, based on the inference regarding the first sensor data, the knowledge base. The first DFRE agent propagates the inference to one or more other DFRE agents in the network.
1. A method comprising:
receiving, at a first deep fusion reasoning engine (DFRE) agent in a network, first sensor data from a first set of one or more sensors in the network;
translating, by the first DFRE agent, the first sensor data into symbolic data;
applying, by the first DFRE agent and using a symbolic knowledge base maintained by the first DFRE agent, symbolic reasoning to the symbolic data to make an inference regarding the first sensor data;
updating, by the first DFRE agent and based on the inference regarding the first sensor data, the symbolic knowledge base;
computing, by the first DFRE agent, a length of a chain of reasoning associated with the inference;
determining, by the first DFRE agent and based on the computed length of the chain of reasoning associated with the inference that the inference should be propagated to one or more other DFRE agents; and
propagating, by the first DFRE agent, the inference to the one or more other DFRE agents in the network.
2. The method as in claim 1 , further comprising:
initiating the symbolic knowledge base with a seed ontology.
3. The method as in claim 1 , wherein propagating the inference to the one or more other DFRE agents in the network comprises:
publishing, by the first DFRE agent, the inference to an in-memory database server in the network, wherein the one or more other DFRE agents subscribe to the server.
4. The method as in claim 1 , wherein the one or more other DFRE agents receive second sensor data from a second set of one or more sensors, and wherein the one or more other DFRE agents analyze the second sensor data using the inference propagated by the first DFRE agent.
5. The method as in claim 1 , further comprising:
identifying, by the first DFRE agent, a class of event associated with the inference; and
determining, by the first DFRE agent, that the inference should be propagated to the one or more other DFRE agents, based on the class of event associated with the inference.
6. The method as in claim 1 , wherein the one or more sensors comprise a camera, and wherein the first sensor data comprises a video stream.
7. The method as in claim 1 , further comprising:
receiving, at the first DFRE agent, an inference made by one of the one or more other DFRE agents in the network; and
updating, by the first DFRE agent and based on the received inference, the symbolic knowledge base maintained by the first DFRE agent.
8. The method as in claim 1 , further comprising:
receiving, at the first DFRE agent, an indication of a focus of attention for the first sensor data, wherein the first DFRE agent applies symbolic reasoning to the symbolic data based in part on the focus of attention.
9. An apparatus, comprising:
one or more network interfaces to communicate with a network;
a processor coupled to the one or more network interfaces and configured to execute one or more processes; and
a memory configured to store a process executable by the processor, the process when executed configured to:
receive first sensor data from a first set of one or more sensors in the network;
translate, by a first DFRE agent executed by the apparatus, the first sensor data into symbolic data;
apply, by the first DFRE agent and using a symbolic knowledge base maintained by the first DFRE agent, symbolic reasoning to the symbolic data to make an inference regarding the first sensor data;
update, by the first DFRE agent and based on the inference regarding the first sensor data, the symbolic knowledge base;
compute, by the first DFRE agent, a length of a chain of reasoning associated with the inference;
determine, by the first DFRE agent and based on the computed length of the chain of reasoning associated with the inference that the inference should be propagated to one or more other DFRE agents;
propagate the inference to the one or more other DFRE agents in the network.
10. The apparatus as in claim 9 , wherein the process when executed is further configured to:
initiate the symbolic knowledge base with a seed ontology.
11. The apparatus as in claim 9 , wherein the apparatus propagates the inference to the one or more other DFRE agents in the network by:
publishing the inference to an in-memory database server in the network, wherein the one or more other DFRE agents subscribe to the server.
12. The apparatus as in claim 9 , wherein the one or more other DFRE agents receive second sensor data from a second set of one or more sensors, and wherein the one or more other DFRE agents analyze the second sensor data using the inference propagated by the first DFRE agent.
13. The apparatus as in claim 9 , wherein the process when executed is further configured to:
identify a class of event associated with the inference; and
determine that the inference should be propagated to the one or more other DFRE agents, based on the class of event associated with the inference.
14. The apparatus as in claim 9 , wherein the one or more sensors comprise a camera, and wherein the first sensor data comprises a video stream.
15. The apparatus as in claim 9 , wherein the process when executed is further configured to:
receive an inference made by one of the one or more other DFRE agents in the network; and
update, based on the received inference, the symbolic knowledge base maintained by the first DFRE agent.
16. The apparatus as in claim 9 , wherein the process when executed is further configured to:
receive an indication of a focus of attention for the first sensor data, wherein the first DFRE agent applies symbolic reasoning to the symbolic data based in part on the focus of attention.
17. A tangible, non-transitory, computer-readable medium storing program instructions that cause a first deep fusion reasoning engine (DFRE) agent in a network to execute a process comprising:
receiving, at the first DFRE agent, first sensor data from a first set of one or more sensors in the network;
translating, by the first DFRE agent, the first sensor data into symbolic data;
applying, by the first DFRE agent and using a symbolic knowledge base maintained by the first DFRE agent, symbolic reasoning to the symbolic data to make an inference regarding the first sensor data;
updating, by the first DFRE agent and based on the inference regarding the first sensor data, the symbolic knowledge base;
computing, by the first DFRE agent, a length of a chain of reasoning associated with the inference;
determining, by the first DFRE agent and based on the computed length of the chain of reasoning associated with the inference that the inference should be propagated to one or more other DFRE agents;
propagating, by the first DFRE agent, the inference to the one or more other DFRE agents in the network.
18. The computer-readable medium as in claim 17 , wherein propagating the inference to the one or more other DFRE agents in the network comprises:
publishing, by the first DFRE agent, the inference to an in-memory database server in the network, wherein the one or more other DFRE agents subscribe to the server.