Method and system for virtual network emulation and self-organizing network control using deep generative models

1. A method performed by a computer device, the method comprising:

training, by the computer device, a generator neural network to simulate a network entity using a discriminator neural network that discriminates output associated with the network entity from output generated by the generator neural network;

receiving, by the computer device, a set of input parameters associated with the simulated network entity;

using, by the computer device, the generator neural network to generate output for the simulated network entity based on the received set of input parameters; and

applying, by the computer device, the generated output for the simulated network entity to manage a communication network.

2. The method of claim 1 , wherein training the generator neural network to simulate the network entity using the discriminator neural network includes:

training the discriminator neural network to classify output associated with the network entity as real data;

training the discriminator neural network to classify output generated by the generator neural network as generated data; and

training the generator neural network to maximize an error associated with the discriminator neural network based on predictions made by the discriminator neural network with respect to the output associated with the network entity and the output generated by the generator neural network.

3. The method of claim 1 , wherein the network entity includes a virtualized network function (VNF) managed object (MO).

4. The method of claim 3 , wherein the VNF MO includes a Long Term Evolution (LTE) wireless access network node.

5. The method of claim 3 , wherein the VNF MO includes at least one of:

an Internet Protocol Multimedia Subsystem (IMS) node,

a router,

a switch,

a firewall,

a gateway, or

a reconfigurable optical add-drop multiplexer (ROADM) controller.

6. The method of claim 3 , further comprising:

determining that a new VNF MO is to be generated;

determining a configuration for the new VNF MO;

estimating a workload for the new VNF MO; and

using the generator neural network to simulate the new VNF MO based on the determined configuration and the estimated workload.

7. The method of claim 3 , further comprising:

training a network performance generator neural network to generate a network performance metric using a network performance discriminator neural network that discriminates output generated by the network performance generator neural network from network performance metrics generated by a plurality of the generator neural networks.

8. The method of claim 7 , further comprising:

receiving a proposed self-organizing network (SON) action;

using the network performance generator neural network to calculate the network performance metric based on the proposed SON action;

determining that the calculated network performance metric satisfies a network performance requirement for the communication network; and

applying the proposed SON action to the communication network, in response to determining that the calculated network performance metric satisfies the network performance requirement for the communication network.

9. The method of claim 8 , further comprising:

training a network model generator neural network to simulate a network model using a discriminator neural network that discriminates a network state from a predicted network state generated by the network model generator neural network.

10. The method of claim 9 , further comprising:

receiving a predicted network state from the network model generator neural network;

receiving a proposed SON action based on the predicted network state;

using the network performance generator neural network to calculate the network performance metric based on the predicted future network state and the proposed SON action; and

updating a SON policy database based on the calculated network performance metric, the predicted network state, and the proposed SON action.

11. The method of claim 1 , wherein the network entity includes an entity configured to generate test cases for testing a performance of the communication network.

12. The method of claim 1 , further comprising:

generating one or more feature vectors for network configurations associated with the generator neural network based on weights associated with nodes of the generator neural network; and

performing configuration vector arithmetic on the one or more feature vectors to determine a particular configuration on the communication network.

13. A computer device comprising:

a memory configured to store instructions; and

a processor configured to execute the instructions to:

train a generator neural network to simulate a network entity using a discriminator neural network that discriminates output associated with the network entity from output generated by the generator neural network;

receive a set of input parameters associated with the simulated network entity;

use the generator neural network to generate output for the simulated network entity based on the received set of input parameters; and

apply the generated output for the simulated network entity to manage a communication network.

14. The computer device of claim 13 , wherein the network entity includes a virtualized network function (VNF) managed object (MO), wherein the VNF MO includes at least one of:

a Long Term Evolution (LTE) wireless access network node,

an Internet Protocol Multimedia Subsystem (IMS) node,

a router,

a switch,

a firewall,

a gateway, or

a reconfigurable optical add-drop multiplexer (ROADM) controller.

15. The computer device of claim 13 , wherein the processor is further configured to execute the instructions to:

determine that a new VNF MO is to be generated;

determine a configuration for the new VNF MO;

estimate a workload for the new VNF MO; and

use the generator neural network to simulate the new VNF MO based on the determined configuration and the estimated workload.

16. The computer device of claim 13 , wherein the processor is further configured to execute the instructions to:

train a network performance generator neural network to generate a network performance metric using a network performance discriminator neural network that discriminates output generated by the network performance generator neural network from network performance metrics generated by a plurality of the generator neural networks.

17. The computer device of claim 16 , wherein the processor is further configured to execute the instructions to:

receive a proposed self-organizing network (SON) action;

use the network performance generator neural network to calculate the network performance metric based on the proposed SON action;

determine that the calculated network performance metric satisfies a network performance requirement for the communication network; and

apply the proposed SON action to the communication network, in response to determining that the calculated network performance metric satisfies the network performance requirement for the communication network.

18. The computer device of claim 17 , wherein the processor is further configured to execute the instructions to:

train a network model generator neural network to simulate a network model using a discriminator neural network that discriminates a network state from a predicted network state generated by the network model generator neural network;

receive a predicted network state from the network model generator neural network;

receive a proposed SON action based on the predicted network state;

use the network performance generator neural network to calculate the network performance metric based on the predicted network state and the proposed SON action; and

update a SON policy database based on the calculated network performance metric, the predicted network state, and the proposed SON action.

19. The computer device of claim 13 , wherein the network entity includes an entity configured to generate test cases for testing a performance of the communication network.

20. A non-transitory computer-readable memory device storing instructions executable by a process, the non-transitory computer-readable memory device comprising:

one or more instructions to train a generator neural network to simulate a network entity using a discriminator neural network that discriminates output associated with the network entity from output generated by the generator neural network;

one or more instructions to receive a set of input parameters associated with the simulated network entity;

one or more instructions to use the generator neural network to generate output for the simulated network entity based on the received set of input parameters; and

one or more instructions to apply the generated output for the simulated network entity to manage a communication network.