JINC: a multi-threaded library for higher-order weighted decision diagram manipulation
نویسنده
چکیده
Ordered Binary Decision Diagrams (OBDDs) have been proven to be an efficient data structure for symbolic algorithms. The efficiency of the symbolic methods depends on the underlying OBDD library. Available OBDD libraries are based on the standard concepts and so far only differ in implementation details. This thesis introduces new techniques to increase run-time and space-efficiency of an OBDD library. This thesis introduces the framework of Higher-Order Weighted Decision Diagrams (HOWDDs) to combine the similarities of different OBDD variants. This framework pioneers the basis for the new variant Toggling Algebraic Decision Diagrams (TADDs) which has been shown to be a space-efficient HOWDD variant for symbolic matrix representation. The concept of HOWDDs has been use to implement the OBDD library JINC. This thesis also analyzes the usage of multi-threading techniques to speed-up OBDD manipulations. A new reordering framework applies the advantages of multi-threading techniques to reordering algorithms. This approach uses an abstraction layer so that the original reordering algorithms are not touched. The challenge that arise from a straight forward algorithm is that the computed-tables and the garbage collection are not as efficient as in a singlethreaded environment. We resolve this problem by developing a new multi-operand APPLY algorithm that eliminates the creation of temporary nodes which could occur during computation and thus reduces the need for caching or garbage collection. The HOWDD framework leads to an efficient library design which has been shown to be more efficient than the established OBDD library CUDD. The HOWDD instance TADD reduces the needed number of nodes by factor two compared to ordinary ADDs. The new multi-threading approaches are more efficient than single-threading approaches by several factors. In the case of the new reordering framework the speedup almost equals the theoretical optimal speed-up. The novel multi-operand APPLY algorithm reduces the memory usage for the n-queens problem by factor 50 which enables the calculation of bigger problem instances compared to the traditional APPLY approach. The new approaches improve the performance and reduce the memory footprint. This leads to the conclusion that applications should be reviewed whether they could benefit from the new multi-threading multi-operand approaches introduced and discussed in this thesis.
منابع مشابه
An Improved Algorithm for Network Reliability Evaluation
Binary Decision Diagram (BDD) is a data structure proved to be compact in representation and efficient in manipulation of Boolean formulas. Using Binary decision diagram in network reliability analysis has already been investigated by some researchers. In this paper we show how an exact algorithm for network reliability can be improved and implemented efficiently by using CUDD - Colorado Univer...
متن کاملA fuzzy multi-objective model for a project management problem
In this research, the multi-objective project management decision problem with fuzzy goals and fuzzy constraints are considered. We constitute α-cut approach and two various fuzzy goal programming solution methods for solving the Multi-Objective Project Management (MOPM) decision problem under fuzzy environments. The Interactive fuzzy multi-objective linear programming (i-FMOLP) and Weighted Ad...
متن کاملSimpler multi-threaded model checking via new foundations for implicit encodings
This research proposes to improve the performance of model checking for asynchronous systems, and the quality and speed of model checking research coding, by orders of magnitude, using the advantages of parallel processing, well defined encodings (GDDs) with desirable properties, enabling clean high-level interfaces to those encodings. Symbolic model checking employs decision diagrams (DDs) to ...
متن کاملTriangular Intuitionistic Fuzzy Triple Bonferroni Harmonic Mean Operators and Application to Multi-attribute Group Decision Making
As an special intuitionistic fuzzy set defined on the real number set, triangular intuitionistic fuzzy number (TIFN) is a fundamental tool for quantifying an ill-known quantity. In order to model the decision maker's overall preference with mandatory requirements, it is necessary to develop some Bonferroni harmonic mean operators for TIFNs which can be used to effectively intergrate the informa...
متن کاملAn open source C++ implementation of multi-threaded Gaussian mixture models, k-means and expectation maximisation
Modelling of multivariate densities is a core component in many signal processing, pattern recognition and machine learning applications. The modelling is often done via Gaussian mixture models (GMMs), which use computationally expensive and potentially unstable training algorithms. We provide an overview of a fast and robust implementation of GMMs in the C++ language, employing multi-threaded ...
متن کامل