The average American consumes over 58 pounds of beef and 141 pounds of milk per year, but cattle farming is a deeply resource-intensive process with significant impacts on land use and carbon emissions – so any gains in efficiency are highly prized. Now, researchers at the University of Florida (UF) have used the university's powerful new HiPerGator AI supercomputer to help ranchers identify the highest-yield livestock. This June, UF made waves when HiPerGator AI, which delivers 17.2 Linpack petaflops, debuted on the Top500 list as the world's third most powerful publicly ranked supercomputer at an educational institution and debuted on the Green500 list as the world's second most efficient publicly ranked supercomputer. So when researchers from UF's Institute of Food and Agricultural Sciences (IFAS) set out to improve cattle yields in a smarter way, they turned to supercomputer-powered AI. "AI has rapidly emerged as a powerful approach in animal genomics and holds great promise to integrate big data from multiple biological layers, leading to accurate prediction of future traits – for example, meat yield," said Raluca Mateescu, a professor of animal science at UF. "My research group is investigating the use of AI methods to develop approaches to accurately predict the value of certain genes. Ultimately, we plan to provide more effective strategies to improve animal productivity."

For a century, researchers have tracked genetic traits to find out which cattle produce more and better milk and meat. Now, two University of Florida scientists will use artificial intelligence to analyze millions of bits of genetic data to try to keep cattle cooler and thus, more productive. Raluca Mateescu, a UF/IFAS professor, and Fernanda Rezende, a UF/IFAS assistant professor – both in animal sciences -- gather hundreds of thousands of pieces of information about cattle genetic traits. They plan to use UF's supercomputer, the HiPerGator, to analyze that data. With the information Mateescu and her team get from the HiPerGator, they can give ranchers better recommendations on which animals to keep and breed for improved quantity of beef and dairy.

The paper investigates parameterized approximate message-passing schemes that are based on bounded inference and are inspired by Pearl's belief propagation algorithm (BP). We start with the bounded inference mini-clustering algorithm and then move to the iterative scheme called Iterative Join-Graph Propagation (IJGP), that combines both iteration and bounded inference. Algorithm IJGP belongs to the class of Generalized Belief Propagation algorithms, a framework that allowed connections with approximate algorithms from statistical physics and is shown empirically to surpass the performance of mini-clustering and belief propagation, as well as a number of other state-of-the-art algorithms on several classes of networks. We also provide insight into the accuracy of iterative BP and IJGP by relating these algorithms to well known classes of constraint propagation schemes.

Inspired by the recently introduced framework of AND/OR search spaces for graphical models, we propose to augment Multi-Valued Decision Diagrams (MDD) with AND nodes, in order to capture function decomposition structure and to extend these compiled data structures to general weighted graphical models (e.g., probabilistic models). We present the AND/OR Multi-Valued Decision Diagram (AOMDD) which compiles a graphical model into a canonical form that supports polynomial (e.g., solution counting, belief updating) or constant time (e.g. equivalence of graphical models) queries. We provide two algorithms for compiling the AOMDD of a graphical model. The first is search-based, and works by applying reduction rules to the trace of the memory intensive AND/OR search algorithm. The second is inference-based and uses a Bucket Elimination schedule to combine the AOMDDs of the input functions via the the APPLY operator. For both algorithms, the compilation time and the size of the AOMDD are, in the worst case, exponential in the treewidth of the graphical model, rather than pathwidth as is known for ordered binary decision diagrams (OBDDs). We introduce the concept of semantic treewidth, which helps explain why the size of a decision diagram is often much smaller than the worst case bound. We provide an experimental evaluation that demonstrates the potential of AOMDDs.