DYNAMIC BOLTZMANN MACHINE FOR ESTIMATING TIME-VARYING SECOND MOMENT

• US 20200134464A1
• Filed: 10/31/2018
• Published: 04/30/2020
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Harry Putra, Rudy Raymond, Osogami, Takayuki, Dasgupta, Sakyasingha
• Granted Patent: US 11,410,042 B2
• Time in Patent Office: Days
• CPC Class Codes: G06N 3/006 based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/047 Probabilistic or stochastic networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06N 3/088 Non-supervised learning, e.g. competitive learning

Abstract

A computer-implemented method includes employing a dynamic Boltzmann machine (DyBM) to predict a higher-order moment of time-series datasets. The method further includes acquiring the time-series datasets transmitted from a source node to a destination node of a neural network including a plurality of nodes, learning, by the processor, a time-series generative model based on the DyBM with eligibility traces, and obtaining, by the processor, parameters of a generalized auto-regressive heteroscedasticity (GARCH) model to predict a time-varying second-order moment of the times-series datasets.

Claims

1. A computer-implemented method executed on a processor for employing a dynamic Boltzmann machine (DyBM) to predict a higher-order moment of time-series datasets, the method comprising:

acquiring the time-series datasets transmitted from a source node to a destination node of a neural network including a plurality of nodes;

learning, by the processor, a time-series generative model based on the DyBM with eligibility traces; and

obtaining, by the processor, parameters of a generalized auto-regressive heteroscedasticity (GARCH) model to predict a time-varying second-order moment of the times-series datasets.

2. The method of claim 1, wherein the time-series datasets are financial time-series data.

3. The method of claim 1, wherein the DyBM is trained online to predict time-varying means by maximizing a log-likelihood that assumes the time-varying second-order moment is fixed.

4. The method of claim 1, wherein a conditional probability density is:

5. The method of claim 4, wherein a Gaussian distribution for each j unit of the conditional probability density is:

6. The method of claim 5, wherein the expected value of the j-th unit is:

7. The method of claim 1, wherein the GARCH is given by:

8. The method of claim 1, wherein the GARCH is given by:

9. A non-transitory computer-readable storage medium comprising a computer-readable program executed on a processor for employing a dynamic Boltzmann machine (DyBM) to predict a higher-order moment of time-series datasets, wherein the computer-readable program when executed on the processor causes a computer to perform the steps of:

acquiring the time-series datasets transmitted from a source node to a destination node of a neural network including a plurality of nodes;

learning, by the processor, a time-series generative model based on the DyBM with eligibility traces; and

obtaining, by the processor, parameters of a generalized auto-regressive heteroscedasticity (GARCH) model to predict a time-varying second-order moment of the times-series datasets.

10. The non-transitory computer-readable storage medium of claim 9, wherein the time-series datasets are financial time-series data.

11. The non-transitory computer-readable storage medium of claim 9, wherein the DyBM is trained online to predict time-varying means by maximizing a log-likelihood that assumes the time-varying second-order moment is fixed.

12. The non-transitory computer-readable storage medium of claim 9, wherein a conditional probability density is:

13. The non-transitory computer-readable storage medium of claim 12, wherein a Gaussian distribution for each j unit of the conditional probability density is:

14. The non-transitory computer-readable storage medium of claim 13, wherein the expected value of the j-th unit is:

15. The non-transitory computer-readable storage medium of claim 9, wherein the GARCH is given by:

16. The non-transitory computer-readable storage medium of claim 9, wherein the GARCH is given by:

17. A system for employing a dynamic Boltzmann machine (DyBM) to predict a higher-order moment of time-series datasets, the system comprising:

a memory; and

one or more processors in communication with the memory configured to;

acquire the time-series datasets transmitted from a source node to a destination node of a neural network including a plurality of nodes;

learn, by the processor, a time-series generative model based on the DyBM with eligibility traces; and

obtain, by the processor, parameters of a generalized auto-regressive heteroscedasticity (GARCH) model to predict a time-varying second-order moment of the times-series datasets.

18. The system of claim 17, wherein the time-series datasets are financial time-series data.

19. The system of claim 17, wherein the GARCH is given by:

20. The system of claim 17, wherein the GARCH is given by: