Learning Tetris using the noisy cross-entropy method

The cross-entropy method is an efficient and general optimization algorithm. However, its applicability in reinforcement learning (RL) seems to be limited because it often converges to suboptimal policies. We apply noise for preventing early convergence of the cross-entropy method, using Tetris, a computer game, for demonstration. The resulting policy outperforms previous RL algorithms by almost two orders of magnitude.     

Research on slope reliability analysis using multi-kernel relevance vector machine and advanced first-order second-moment method

Engineering with Computers - To increase the efficiency and accuracy in slope stability analysis, a reliability analysis method based on machine learning and the advanced first-order second-moment...     

A first-order multigrid method for bound-constrained convex optimization

The aim of this paper is to design an efficient multigrid method for constrained convex optimization problems arising from discretization of some underlying infinite dimensional problems. Due to problem dependency of this approach, we only consider bound constraints with (possibly) a single equality constraint. As our aim is to target large-scale problems, we want to avoid computation of second derivatives of the objective function, thus excluding Newton-like methods. We propose a smoothing operator that only uses first-order information and study the computational efficiency of the resulting method.     

Coupling the cross-entropy with the line sampling method for risk-based design optimization

An algorithm for risk-based optimization (RO) of engineering systems is proposed, which couples the Cross-entropy (CE) optimization method with the Line Sampling (LS) reliability method. The CE-LS algorithm relies on the CE method to optimize the total cost of a system that is composed of the design and operation cost (e.g., production cost) and the expected failure cost (i.e., failure risk). Guided by the random search of the CE method, the algorithm proceeds iteratively to update a set of random search distributions such that the optimal or near-optimal solution is likely to occur. The LS-based failure probability estimates are required to evaluate the failure risk. Throughout the optimization process, the coupling relies on a local weighted average approximation of the probability of failure to reduce the computational demands associated with RO. As the CE-LS algorithm proceeds to locate a region of design parameters with near-optimal solutions, the local weighted average approximation of the probability of failure is refined. The adaptive refinement procedure is repeatedly applied until convergence criteria with respect to both the optimization and the approximation of the failure probability are satisfied. The performance of the proposed optimization heuristic is examined empirically on several RO problems, including the design of a monopile foundation for offshore wind turbines.     

利用优化的MODBUS协议实现分布式控制

MODBUS协议是控制领域应用非常广泛的协议。对标准的MODBUS协议在多任务异步处理中的局限性进行了研究,并在此基础上提出了一种基于物理地址标记的异步MODBUS协议------a-MODBUS协议。该协议利用物理地址功能的相对确定性,进行报文级任务逻辑功能识别。基于此协议,提出了一种基于agent的任务调度模型,并且在工程实践中得到成功应用。     

Improvement of turbo decoding using cross-entropy

In this paper, we present a simple method for normalizing the output information produced by a turbo decoder. The method is devised based on the cross-entropy (CE) concept. Simulations comparing the new method with some widely used normalization techniques show that the proposed approach can achieve about 0.2～0.3 dB coding gain improvement on average while reducing up to about 1/2～2/3 iteration for decoding, but require much fewer and simpler computations.     

软件系统可靠性优化的一种改进方法研究

基于并行遗传算法将软件系统的可靠性优化问题表达为一类带约束条件的组合优化问题,并采用并行遗传算法中的岛屿模型和迁移策略,较好地改善了搜索性能。模拟实验表明:并行遗传算法有效地提高了运行速度和求解质量。     

An approximate sequential optimization and reliability assessment method for reliability-based design optimization

Sequential optimization and reliability assessment (SORA) is one of the most popular decoupled approaches to solve reliability-based design optimization (RBDO) problem because of its efficiency and robustness. In SORA, the double loop structure is decoupled through a serial of cycles of deterministic optimization and reliability assessment. In each cycle, the deterministic optimization and reliability assessment are performed sequentially and the boundaries of violated constraints are shifted to the feasible direction according to the reliability information obtained in the previous cycle. In this paper, based on the concept of SORA, approximate most probable target point (MPTP) and approximate probabilistic performance measure (PPM) are adopted in reliability assessment. In each cycle, the approximate MPTP needs to be reserved, which will be used to obtain new approximate MPTP in the next cycle. There is no need to evaluate the performance function in the deterministic optimization since the approximate PPM and its sensitivity are used to formulate the linear Taylor expansion of the constraint function. One example is used to illustrate that the approximate MPTP will approach the accurate MPTP with the iteration. The design variables and the approximate MPTP converge simultaneously. Numerical results of several examples indicate the proposed method is robust and more efficient than SORA and other common RBDO methods.     

Refined modeling of sensor reliability in the belief function framework using contextual discounting

In belief functions theory, the discounting operation allows to combine information provided by a source in the form of a belief function with meta-knowledge regarding the reliability of that source, resulting in a “weakened”, less informative belief function. In this article, an extension of the discounting operation is proposed, allowing to use more detailed information regarding the reliability of the source in different contexts, i.e., conditionally on different hypotheses regarding the variable on interest. This results in a contextual discounting operation parameterized with a discount rate vector. Some properties of this contextual discounting operation are studied, and its relationship with classical discounting is explained. A method for learning the discount rates is also presented.     

A machine tool matching method in cloud manufacturing using Markov Decision Process and cross-entropy

Cloud Manufacturing (CMfg) is a state-of-the-art manufacturing paradigm implementing the concept of service-oriented manufacturing. Machine tools are one kind of the critical manufacturing resources in Cloud Manufacturing, however machine tool matching is still immature owning to customization manufacturing service demands from users and various disturbing factors in production. This paper proposes a machine tool matching method for dealing with a single Cloud Manufacturing task with complex machine tool application demands. In this method, the demands of machine tools and themselves are described and evaluated based on a universal framework to obtain candidate resource groups satisfying local requirements of sub-demands. Then, a series of Markov Decision Processes (MDP) is established, which take the minimal service cost as optimal object to meet global requirements, and a cross-entropy based algorithm is used to solve the optimal object. Finally, simulation experiments are conducted to validate the usability and superiority in efficiency of the proposed method.     

基于交叉熵的ITFN的多属性决策方法

针对属性值为直觉梯形模糊数且属性权重完全未知的多属性决策问题,提出了一种基于交叉熵的决策方法。给出期望值的方法将直觉梯形模糊数转化为直觉模糊数,进而提出直觉模糊数的交叉熵等概念及相关性质。基于各方案与正理想方案的总区别信息最小化原则,建立非线性模型,求出属性权重。用实例说明该方法的有效性。     

An accelerated first-order method for solving SOS relaxations of unconstrained polynomial optimization problems

Our interest lies in solving sum of squares (SOS) relaxations of large-scale unconstrained polynomial optimization problems. Because interior-point methods for solving these problems are severely limited by the large-scale, we are motivated to explore efficient implementations of an accelerated first-order method to solve this class of problems. By exploiting special structural properties of this problem class, we greatly reduce the computational cost of the first-order method at each iteration. We report promising computational results as well as a curious observation about the behaviour of the first-order method for the SOS relaxations of the unconstrained polynomial optimization problem.     

A new method for reliability analysis and reliability-based design optimization

Structural and Multidisciplinary Optimization - We present a novel method for reliability-based design optimization, which is based on the approximation of the safe region in the random space by a...     

An importance learning method for non-probabilistic reliability analysis and optimization

With the time-consuming computations incurred by nested double-loop strategy and multiple performance functions, the enhancement of computational efficiency for the non-probabilistic reliability estimation and optimization is a challenging problem in the assessment of structural safety. In this study, a novel importance learning method (ILM) is proposed on the basis of active learning technique using Kriging metamodel, which builds the Kriging model accurately and efficiently by considering the influence of the most concerned point. To further accelerate the convergence rate of non-probabilistic reliability analysis, a new stopping criterion is constructed to ensure accuracy of the Kriging model. For solving the non-probabilistic reliability-based design optimization (NRBDO) problems with multiple non-probabilistic constraints, a new active learning function is further developed based upon the ILM for dealing with this problem efficiently. The proposed ILM is verified by two non-probabilistic reliability estimation examples and three NRBDO examples. Comparing with the existing active learning methods, the optimal results calculated by the proposed ILM show high performance in terms of efficiency and accuracy.

     

Optimal fuzzy control system using the cross-entropy method. A case study of a drilling process

This paper focuses on the optimal tuning of fuzzy control systems using the cross-entropy precise mathematical framework. The design of an optimal fuzzy controller for cutting force regulation in a network-based application and applied to the drilling process is described. The key issue is to obtain optimal fuzzy controller parameters that yield a fast and accurate response with minimum overshoot by minimising the integral time absolute error (ITAE) performance index. Simulation results show that the cross-entropy method does find the optimal solution (i.e. input scaling factors) very accurately, and it can be programmed and implemented very easily (few setting parameters). The results of a comparative study demonstrate that optimal tuning with the cross-entropy method provides a good transient response (without overshoot) and a better error-based performance index than simulated annealing [17], the Nelder-Mead method [14] and genetic algorithms [33]. The experimental results demonstrate that the proposed optimal fuzzy control provides outstanding transient response without overshoot, a small settling time and a minimum steady-state error. The application of optimal fuzzy control reduces rapid drill wear and catastrophic drill breakage due to the increasing and oscillatory cutting forces that occur as the drill depth increases.     

Derivation of regression formulae using the minimum cross-entropy principle

We derive expressions for linear regression directly by minimizing cross-entropy subject to the observational evidence. We also show that in the continuous case the MEP (minimum cross-entropy principle) reproduces relative frequency evidence a posteriori. These calculations support the exclusive use of the MEP for inductive inference.     

Numerical method for shape optimization using meshfree method

A numerical method for continuum-based shape design sensitivity analysis and optimization using the meshfree method is proposed. The reproducing kernel particle method is used for domain discretization in conjunction with the Gauss integration method. Special features of the meshfree method from a sensitivity analysis viewpoint are discussed, including the treatment of essential boundary conditions, and the dependence of the shape function on the design variation. It is shown that the mesh distortion that exists in the finite element-based design approach is effectively resolved for large shape changing design problems through 2-D and 3-D numerical examples. The number of design iterations is reduced because of the accurate sensitivity information.     

Global optimization using a multipoint type quasi-chaotic optimization method

This paper proposes a new global optimization method called the multipoint type quasi-chaotic optimization method. In the proposed method, the simultaneous perturbation gradient approximation is introduced into a multipoint type chaotic optimization method in order to carry out optimization without gradient information. The multipoint type chaotic optimization method, which has been proposed recently, is a global optimization method for solving unconstrained optimization problems in which multiple search points which implement global searches driven by a chaotic gradient dynamic model are advected to their elite search points (best search points among the current search histories). The chaotic optimization method uses a gradient to drive search points. Hence, its application is restricted to a class of problems in which the gradient of the objective function can be computed. In this paper, the simultaneous perturbation gradient approximation is introduced into the multipoint type chaotic optimization method in order to approximate gradients so that the chaotic optimization method can be applied to a class of problems for which only the objective function values can be computed. Then, the effectiveness of the proposed method is confirmed through application to several unconstrained multi-peaked, noisy, or discontinuous optimization problems with 100 or more variables, comparing to other major meta-heuristics.     

Coordinative optimization method of composite laminated structures based on system reliability

This paper develops the coordinative optimization method based on system reliability for laminated structures. The proposed method improves the rough RBO based on first layer failure (FLF) criterion for composite laminates, and the coupling optimization method of thickness and sequence in traditional RBO strategy based on last layer failure criterion (LLF) is improved. In this paper, the finite element analysis is used to obtain the response for the failure based on two-dimension Hashin failure criterion (the limit function). Obviously, the stiffness of composite materials will decline due to destruction of elements. Therefore, stiffness degradation is considered to describe the process of damage evolution. Subsequently, combining with the branch-bound method (B&B), we can complete the search of main failure sequences and calculate the system reliability with the help of the second-order upper bound theory. In order to guarantee the efficiency and accuracy of optimization, the adaptive GA algorithm is introduced in the whole optimization procedure. After the proposed optimization policy is given in detail, two laminated structures are presented and the results are compared with the traditional optimal method based on safety factor, which demonstrates the validity and reasonability of the developed methodology.