Stochastic search in data-based modelling of dynamic systems

Experiments with new stochastic global optimization search techniques

Although stochastic search techniques have shown promise in test generation and design validation, they often fail when there is a specific, random-resistant sequence of vectors required to exercise a target. In order to combat this, deterministic techniques are added, resulting in a hybrid solution to maintain high speed of execution while improving metric performance. This paper presents a formal hybridization that combines a Register Transfer Level (RTL) stochastic swarm intelligence based test vector generation with the Verilator Verilog-to-C++ source-to-source compiler. Verilator generates a fast cycle accurate C++ simulation unit for Verilog descriptions and provides instrumentation for branch and toggle coverage metrics. This RTL model can also be used to generate a bounded model checking (BMC) instance. During the stochastic search, the bounded model checker is launched to expand the unexplored search frontier and aid in the navigation of narrow paths. Additionally, an inductive reach ability test is applied in order to eliminate unreachable branches from our search space. These additions have significantly improved branch coverage, reaching 100% in several ITC99 benchmarks. Additionally, compared to previous functional test generation methods, we achieve substantial speedup achieved with purely stochastic methods.

Quality is the hallmark of a competitive product. It is necessary to use inspection stations to check product quality and process performance. In this paper, the authors are concerned with the problem of location of inspection stations in a multistage manufacturing system. The authors present two stochastic search algorithms for solving this problem, one based on simulated annealing and the other on genetic algorithms. These algorithms are developed to determine the location of inspection stations resulting in a minimum expected total cost in a multistage manufacturing system. The total cost includes inspection, processing and scrapping cost at each stage of the production process. A penalty cost is also included in it to account for a defective item which is not detected by the inspection scheme. A set of test examples are solved using these algorithms. The authors also compare performance of these two algorithms.

It is important to determine supporting locations for structural stability when a structure is loaded with non-uniform load or supporting locations as well as the number of the supporting structures are restricted by the problem of space. Moreover, the supporting location optimization of complex structure in real world is fTequently faced with discontinuous design space. Therefore, the traditional optimization methods based on derivative are not suitable Whereas, Genetic Algorithm (GA) based on stochastic search technique is a very robust and general method. The KSTAR in-vessel control coil installed in vacuum vessel is loaded with nonelectro-magnetic load and supporting locations are restricted by the problem of space. This paper shows the supporting location optimization for structural stability of the in-vessel control coil. Optimization has been performed by means of a developed program. It consists of a Finite Element Analysis interfaced with a Genetic Algorithm. In addition, this paper presents an algorithm to find an optimum solution in discontinuous space using continuous design variables.

Real-time embedded systems must operate reliably and predictably from both a functional and dynamic point of view. To achieve these objectives we need to use an integrated approach to the design of systems. Integrated, that is, in terms of designing the hardware and software as a co-design operation. We present a co-design tool for the evaluation and demonstration of real-time system architectures and performance tradeoffs. The tool enables simulation and animation techniques to be applied in the analysis and early design stages of a project. It is specifically intended for use with multiprocessor systems to determine the optimum task allocation strategy. Allocations must consider task priority, timing characteristics and node utilization. This is achieved by the application of a stochastic global search technique based on genetic algorithm (GA) principles. Solutions produced by the genetic algorithm can be simulated and presented graphically to allow users to more easily estimate the quality of the allocation strategy. >

The main objective of this paper is to provide a state-of-the-art review, analyze and discuss stochastic local search techniques used for solving hard combinatorial problems. It begins with a short introduction, motivation and some basic notation on combinatorial problems, search paradigms and other relevant features of searching techniques as needed for background. In the following a brief overview of the stochastic local search methods along with an analysis of the state-of-the-art stochastic local search algorithms is given. Finally, the last part of the paper present and discuss some of the most latest trends in application of stochastic local search algorithms in machine learning, data mining and some other areas of science and engineering. We conclude with a discussion on capabilities and limitations of stochastic local search algorithms.

Although the later stages of design are well supported by software, there is currently no commercially available software to support conceptual design. The research community has been aware of this deficiency and has been working hard to establish techniques which are capable of assisting the designer to find better solutions at the conceptual design stage. Within the last few years, a set of algorithms known collectively as stochastic search techniques, have emerged as being ideal tools for this area. This paper describes how one of these algorithms, known as a genetic algorithm, works and how it can be used to support the designer during conceptual design.

In this paper, a stochastic search technique, namely Particle Swarm Optimization (PSO) is used to determine the optimal locations of two Static Var Compensators (SVCs) in an IEEE 14-bus system. Static Var compensators (SVC) are the most widely used shunt FACTS devices within power networks because of their smaller costs and significant system enhancements. Appeared about two decades ago, the SVC is mainly installed for voltage support and furthermore, when installed in a proper location, it can reduce the power loss. Here the problem is framed as an optimization task and the optimal locations of SVC are identified using the novel technique. The efficacy of the new algorithm is tested with extensive computer simulations and further compared with Genetic Algorithm (GA) based approach. The performance analysis is done through extensive simulations and shows that the proposed dispensation is on a par with existing techniques.

This paper has proposed a distributed power control algorithm that uses a stochastic search technique in order to solve linear systems of equations for power update in CDMA cellular radio systems. The proposed algorithm is developed by applying Bremermann's evolutionary computation algorithm to the CDMA power control problem. The major gain from the applied evolutionary computation algorithm is more rapid optimization on linear systems of equations compared with the simple genetic algorithm (SGA). Employing the distributed constrained power control (DCPC) as a reference algorithm, we have designed and implemented computational experiments on the DS-CDMA system. The results indicate that the proposed algorithm significantly enhances the optimization speed of power control. The proposed algorithm is also compared with the bang-bang type algorithm used in the IS-95 and the W-CDMA systems. The results show that our proposed algorithm also has a high potential advantage for increasing the CDMA cellular radio network capacity and decreasing the mobile terminal power consumption.

Impact force location in frame structures

Program synthesis involves discovering a program from an underlying space of programs that satisfies a given specification using some search technique.3 It has many applications including algorithm discovery, optimized implementations, programming assistance,5 and synthesis of small scripts to automate repetitive tasks for end users.4 Its success relies heavily on efficient search algorithms to navigate the underlying huge state space of programs. The authors of the following paper have developed a stochastic search technique and applied it to program optimization. The impressive results of their implementation STOKE on hard program optimization benchmarks illustrate the promising potential of stochastic search to hard program synthesis problems.

Learning MAX-SAT from contextual examples for combinatorial optimisation

Evolutionary algorithms (EA), such as the genetic algorithm (GA), have found wide popularity because of their capability of global optimum search. As an answer to the quest for faster converging evolutionary algorithms, particle swarm optimization (PSO) was introduced to the electromagnetics domain (Robinson, J. and Rahmat-Samii, Y., IEEE Trans. Antennas Propag., vol.AP-52, no.2, p.397-407, 2004). PSO is a relatively new stochastic evolutionary computation technique, based on the movement and intelligence of swarms. PSO is found to hold a superior position with respect to other evolutionary methods, such as GA. Since PSO is a very recent method, many of its capabilities are still unexplored. It benefits from the inherent simplicity of GA while it converges much faster than conventional evolutionary algorithms. The paper presents a simple and universal design procedure for planar microwave filters. To accelerate the computation phase, an equivalent circuit topology is derived from FDTD formulation. The method is fast and robust and is well suited to the fast paced engineering design environment. As a case study, a microwave band-pass filter is designed.

Since 1991, tries were made to enhance the stochastic local search techniques (SLS). Some researchers turned their focus on studying the structure of the propositional satisfiability problems (SAT) to better understand their complexity in order to come up with better algorithms. Other researchers focused in investigating new ways to develop heuristics that alter the search space based on some information gathered prior to or during the search process. Thus, many heuristics, enhancements and developments were introduced to improve SLS techniques performance during the last three decades. As a result a group of heuristics were introduced namely Dynamic Local Search (DLS) that could outperform the systematic search techniques. Interestingly, a common characteristic of DLS heuristics is that they all depend on the use of weights during searching for satisfiable formulas. In our study we experimentally investigated the weights behaviors and movements during searching for satisfiability using DLS techniques, for simplicity, DDFW DLS heuristic is chosen. As a results of our studies we discovered that while solving hard SAT problems such as blocks world and graph coloring problems, weights stagnation occur in many areas within the search space. We conclude that weights stagnation occurrence is highly related to the level of the problem density, complexity and connectivity.

Variable step search algorithm for feedforward networks