Bayesian Network Generation via Functional Dependency Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for learning Bayesian networks face challenges in accurately capturing causal relationships due to limitations on the number of parents per node, which can lead to functional dependencies being overlooked, resulting in an incomplete representation of data distributions.
Innovation Solution
A method that identifies and exploits functional dependencies between variables by determining minimal parent sets and constructing a directed acyclic graph, allowing for the accurate representation of causal relationships while maintaining a manageable network size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the number of parents per node is limited to control network size, then the network size is reduced and becomes manageable, but functional dependencies cannot be learned and causal relationships are incompletely captured
Solution Approach 1:
The patent extracts functional dependencies from the data as separate structural constraints before learning the Bayesian network. By identifying and removing variables that are functionally determined by others, the method reduces the effective network size while preserving all causal relationships. This extraction step prevents the loss of functional dependencies that would otherwise be truncated by parent number limits.
Solution Approach 2:
The patent performs preliminary analysis to identify functional dependencies in the data before the main Bayesian network learning process. By pre-processing the data to detect deterministic relationships and removing redundant variables upfront, the method ensures that all causal relationships are captured before the network structure learning begins, avoiding the need to later discover relationships that were truncated by size constraints.
2Productivity
If restrictions are imposed on learnable structures to make learning feasible, then the learning task becomes algorithmically manageable, but the ability to accurately capture causal relationships is reduced
Solution Approach 1:
The patent extracts functional dependencies as separate structural constraints before learning the Bayesian network. By identifying and removing variables that are functionally determined by others, the method reduces the effective network size while preserving all causal relationships. This extraction step prevents the loss of functional dependencies that would otherwise be truncated by parent number limits.
Solution Approach 2:
The patent performs preliminary analysis to identify functional dependencies in the data before the main Bayesian network learning process. By pre-processing the data to detect deterministic relationships and removing redundant variables upfront, the method ensures that all causal relationships are captured before the network structure learning begins, avoiding the need to later discover relationships that were truncated by size constraints.
Data Source
AI summary
A computer implemented method, computer program product and system for generating a Bayesian network. A dataset comprising multiple instances of multiple variables is received. A target variable from the received dataset is selected. Multiple parent sets of variables for the target variable are determined, such that, for each parent set of variables, the target variable is functionally dependent on the respective parent set of variables. For multiple variables of the received dataset, the selecting of a new target variable from the received dataset and determining multiple parent sets of variables for the new target variable is repeated. A Bayesian network (includes a directed acyclic graph of nodes and edges) is then generated for the variables such that one or more of the determined parent sets of variables for the target variables are inserted into the graph and edges from the graph are removed to ensure that the graph is acyclic.