The name ★-IPS (star-IPS) covers a family of algorithms, based on Iterative Proportional Scaling, intended for Gaussian Markov random field selection. ★-IPS is described in the following article:
Martin V., Furtlehner C., Han Y., and Lasgouttes J.-M., GMRF Estimation under Topological and Spectral constraints, presented at ECML/PKDD 2014, Nancy.
Abstract: We investigate the problem of Gaussian Markov random field selection under a non-analytic constraint: the estimated models must be compatible with a fast inference algorithm, namely the Gaussian belief propagation algorithm. To address this question, we introduce the ★-IPS framework, based on iterative proportional scaling, which incrementally selects candidate links in a greedy manner. Besides its intrinsic sparsity-inducing ability, this algorithm is flexible enough to incorporate various spectral constraints, like e.g. walk summability, and topological constraints, like avoiding the formation of short loops. Experimental tests on various datasets, including traffic data from San Fransisco Bay area, indicate that this approach can deliver, with reasonable computational cost, a broad range of efficient inference models, which are not accessible through penalization with traditional sparsity-inducing norms.
The source code for our implementation of the algorithm is available on the Inria Gitlab ★-IPS repository.
Version 1.3 is available here (June 25, 2019).
Our git repository is accessible using one of the following commands:
git clone https://gitlab.inria.fr/bptraffic/star-ips.git
git clone email@example.com:bptraffic/star-ips.git
For any request about ★-IPS framework or our implementation, please contact Cyril Furtlehner.