Generating Different Sampling Orders of Random Variables in a Bayesian Model for Markov Chain Monte Carlo Sampling Techniques

Patent metadata

Field	Value
Publication number	US-2025111257-A1
Application number	US-202318478822-A
Country	US
Kind code	A1
Filing date	Sep 29, 2023
Priority date	Sep 29, 2023
Publication date	Apr 3, 2025
Grant date	—

Abstract

Different sampling orders of random variables in a Bayesian model may be generated for Markov Chain Monte Carlo sampling techniques. Code may be received that causes a Markov Chain Monte Carlo sampling technique to be performed with respect to a Bayesian model that includes random variables representing different parameterized probability distributions and connected via edges in a Directed Acyclical Graph (DAG). Instructions may be generated to execute the code that cause the Markov Chain Monte Carlo sampling technique, the instructions including performing different orders for sampling different random variables in the DAG in different iterations of the Markov Chain Monte Carlo sampling technique.

First claim

What is claimed:

1 . A system, comprising:
at least one processor; a memory, comprising program instructions that when executed by the at least one processor cause the at least one processor to implement a system configured to:
receive code that causes a Markov Chain Monte Carlo sampling technique to be performed with respect to a Bayesian model comprising a plurality of random variables, representing different parameterized probability distributions and connected via edges in a directed acyclical graph; and
evaluate the directed acyclical graph to determine a first order for sampling from the plurality of random variables for a first iteration of the Markov Chain Monte Carlo sampling technique and performing a second order, different than the first order, for sampling from the plurality of random variables for a second iteration of the Markov Chain Monte Carlo sampling technique;
generate instructions to execute the code that causes the Markov Chain Monte Carlo sampling technique, wherein the instructions comprise performing the first order and the second order for different respective iterations of the Markov Chain Monte Carlo sampling technique; and
perform the instructions to execute the code.

2 . The system of claim 1 , wherein the Markov Chain Monte Carlo sampling technique is a Gibbs sampling technique.

3 . The system of claim 1 , wherein the second order is a reverse of the first order.

4 . The system of claim 1 , wherein the code that causes the Markov Chain Monte Carlo sampling technique to be performed also generates an inference for a given input value.

5 . The system of claim 1 , wherein the first order is performed on odd numbered iterations of the Markov Chain Monte Carlo sampling technique and wherein the second order is performed on even numbered iterations of the Markov Chain Monte Carlo sampling technique.

6 . The system of claim 1 , wherein the Markov Chain Monte Carlo sampling technique is a Metropolis Hastings sampling technique.

7 . A method, comprising:
performing, by one or more computing devices:
receiving code that causes a Markov Chain Monte Carlo sampling technique to be performed with respect to a Bayesian model comprising a plurality of random variables, representing different parameterized probability distributions; and
generating instructions to execute the code that causes the Markov Chain Monte Carlo sampling technique, wherein the instructions comprise performing a first order for sampling from the plurality of random variables for a first iteration of the Markov Chain Monte Carlo sampling technique and performing a second order, different than the first order, for sampling from the plurality of random variables for a second iteration of the Markov Chain Monte Carlo sampling technique.

8 . The method of claim 7 , wherein the Markov Chain Monte Carlo sampling technique is a Gibbs sampling technique.

9 . The method of claim 7 , wherein the second order is a reverse of the first order.

10 . The method of claim 7 , wherein the code that causes the Markov Chain Monte Carlo sampling technique to be performed generates an inference for a given input value.

11 . The method of claim 7 , wherein the first order is performed on odd numbered iterations of the Markov Chain Monte Carlo sampling technique and wherein the second order is performed on even numbered iterations of the Markov Chain Monte Carlo sampling technique.

12 . The method of claim 7 , wherein the code is specified in a probabilistic programming language and wherein the instructions are generated in a non-probabilistic programming language.

13 . The method of claim 7 , wherein the Markov Chain Monte Carlo sampling technique is a Metropolis Hastings sampling technique.

14 . One or more non-transitory, computer-readable storage media, storing program instructions that when executed on or across one or more computing devices, cause the one or more computing devices to implement:
receiving code that causes a Markov Chain Monte Carlo sampling technique to be performed with respect to a Bayesian model comprising a plurality of random variables, representing different parameterized probability distributions; and evaluating the plurality of random variables of the Bayesian model to determine a first order for sampling from the plurality of random variables for a first iteration of the Markov Chain Monte Carlo sampling technique and performing a second order, different than the first order, for sampling from the plurality of random variables for a second iteration of the Markov generating instructions to execute the code that causes the Markov Chain Monte Carlo sampling technique, wherein the instructions comprise performing the first order and the second order for different respective iterations of the Markov Chain Monte Carlo sampling technique.

15 . The one or more non-transitory, computer-readable storage media of claim 14 , wherein the Markov Chain Monte Carlo sampling technique is a Gibbs sampling technique.

16 . The one or more non-transitory, computer-readable storage media of claim 14 , wherein the second order is a reverse of the first order.

17 . The one or more non-transitory, computer-readable storage media of claim 14 , wherein the code that causes the Markov Chain Monte Carlo sampling technique to be performed generates an inference for a given input value.

18 . The one or more non-transitory, computer-readable storage media of claim 14 , wherein the first order is performed on odd numbered iterations of the Markov Chain Monte Carlo sampling technique and wherein the second order is performed on even numbered iterations of the Markov Chain Monte Carlo sampling technique.

19 . The one or more non-transitory, computer-readable storage media of claim 14 , wherein the code is specified in a probabilistic programming language and wherein the instructions are generated in a non-probabilistic programming language.

20 . The one or more non-transitory, computer-readable storage media of claim 14 , wherein the Markov Chain Monte Carlo sampling technique is a Metropolis Hastings sampling technique.

Assignees

• Oracle Int Corp

Inventors

• Goodman Daniel

Classifications

• G06N7/01 Primary

  Probabilistic graphical models, e.g. probabilistic networks · CPC title