joint pseudolikelihood
Definition
Joint pseudolikelihood is an approximation method for estimating parameters of the Ising model that replaces the exact likelihood with a product of conditional distributions of each variable given all remaining variables. Unlike the exact likelihood, which requires evaluation of an intractable normalizing constant with exponentially growing terms, the joint pseudolikelihood uses normalization constants with only linearly scaling terms, making it computationally tractable for larger networks. Maximum pseudolikelihood estimation based on the joint pseudolikelihood has been shown to yield consistent estimators and serves as a stable method for approximating maximum likelihood estimates, particularly performing well for dense networks and when paired with regularization approaches for edge selection.
Sources: Keetelaar et al. (2024)
Related Terms
Applications
Joint Pseudolikelihood and the Ising Model
Joint pseudolikelihood is fundamentally designed as an approximation method for the Ising model. The joint pseudolikelihood approach addresses the computational intractability of exact maximum likelihood estimation for the Ising model by circumventing direct evaluation of the normalizing constant.
Sources: Keetelaar et al. (2024)
Joint Pseudolikelihood and Disjoint Pseudolikelihood
Joint pseudolikelihood and disjoint pseudolikelihood are two competing approximation methods for likelihood-based estimation of the Ising model, each with different computational and performance characteristics. Maximum pseudolikelihood estimation based on joint pseudolikelihood demonstrates greater stability and better performance for dense networks, whereas disjoint pseudolikelihood performs better for large sample sizes and sparse networks.
Sources: Keetelaar et al. (2024)
Joint Pseudolikelihood and Edge Selection
Joint pseudolikelihood paired with regularization has been found to outperform disjoint pseudolikelihood approaches for edge selection in network models, suggesting that the approximation properties of the joint pseudolikelihood are particularly advantageous when combined with regularization techniques.
Sources: Keetelaar et al. (2024)
