Pruning Isomorphic Structural Sub-problems in Configuration
Stephane Grandcolas, Laurent Henocque, Nicolas Prcovic
备注:
This research report contains the proofs and full details missing from the short paper "A Canonicity Test for Configuration" in proceedings of conference CP'03
Configuring consists in simulating the realization of a complex product from a catalog of component parts, using known relations between types, and picking values for object attributes. This highly combinatorial problem in the field of constraint programming has been addressed with a variety of approaches since the foundation system R1(McDermott82). An inherent difficulty in solving configuration problems is the existence of many isomorphisms among interpretations. We describe a formalism independent approach to improve the detection of isomorphisms by configurators, which does not require to adapt the problem model. To achieve this, we exploit the properties of a characteristic subset of configuration problems, called the structural sub-problem, which canonical solutions can be produced or tested at a limited cost. In this paper we present an algorithm for testing the canonicity of configurations, that can be added as a symmetry breaking constraint to any configurator. The cost and efficiency of this canonicity test are given.
One-Cycle Structured Pruning via Stability-Driven Subnetwork Search
IT Application Research Center, Korea Electronics Technology Institute(韩国电子技术研究所应用信息技术中心);Department of Intelligent Semiconductors, Soongsil University(顺慈大学智能半导体系);Department of Computer Science & Engineering, Seoul National University(首尔国立大学计算机科学与工程系)
All-in-One Tuning and Structural Pruning for Domain-Specific LLMs
Northeastern University(东北大学);George Mason University(乔治梅森大学);GE HealthCare(GE医疗);University of Pennsylvania(宾夕法尼亚大学);University of Pittsburgh(匹兹堡大学);University of Florida(佛罗里达大学);Florida State University(佛罗里达州立大学)