A single-machine replacement model with learning

The replacement or upgrade of productive resources over time is an important decision for a manufacturing organization. The type of technology used in the productive resources determines how effectively the manufacturing operations can support the product and marketing strategy of the organization. Increasing operating costs (cost of maintenance, labor, and depreciation) over time force manufacturing organizations to periodically consider replacement or upgrade of their existing productive resources. We assume that there is a setup cost associated with the replacement of a machine, and that the setup cost is a nonincreasing function of the number of replacements made so far due to learning in setups. The operating cost of a newer machine is assumed to be lower than the operating cost of an older machine in any given period, except perhaps in the first period of operation of the new machine when the cost could be unusually high due to higher initial depreciation. A forward dynamic programming algorithm is developed which can be used to solve finite-horizon problems. We develop procedures to find decision and forecast horizons such that choices made during the decision horizon based only on information over the forecast horizon are also optimal for any longer horizon problem. Thus, we are able to obtain optimal results for what is effectively an infinite-horizon problem while only requiring data over a finite period of time. We present a numerical example to illustrate the decision/forecast horizon procedure, as well as a study of the effects of considering learning in making a series of machine replacement decisions. © 1993 John Wiley & Sons. Inc.     

Allocation of quality improvement targets based on investments in learning

Purchased materials often account for more than 50% of a manufacturer's product nonconformance cost. A common strategy for reducing such costs is to allocate periodic quality improvement targets to suppliers of such materials. Improvement target allocations are often accomplished via ad hoc methods such as prescribing a fixed, across‐the‐board percentage improvement for all suppliers, which, however, may not be the most effective or efficient approach for allocating improvement targets. We propose a formal modeling and optimization approach for assessing quality improvement targets for suppliers, based on process variance reduction. In our models, a manufacturer has multiple product performance measures that are linear functions of a common set of design variables (factors), each of which is an output from an independent supplier's process. We assume that a manufacturer's quality improvement is a result of reductions in supplier process variances, obtained through learning and experience, which require appropriate investments by both the manufacturer and suppliers. Three learning investment (cost) models for achieving a given learning rate are used to determine the allocations that minimize expected costs for both the supplier and manufacturer and to assess the sensitivity of investment in learning on the allocation of quality improvement targets. Solutions for determining optimal learning rates, and concomitant quality improvement targets are derived for each learning investment function. We also account for the risk that a supplier may not achieve a targeted learning rate for quality improvements. An extensive computational study is conducted to investigate the differences between optimal variance allocations and a fixed percentage allocation. These differences are examined with respect to (i) variance improvement targets and (ii) total expected cost. For certain types of learning investment models, the results suggest that orders of magnitude differences in variance allocations and expected total costs occur between optimal allocations and those arrived at via the commonly used rule of fixed percentage allocations. However, for learning investments characterized by a quadratic function, there is surprisingly close agreement with an “across‐the‐board” allocation of 20% quality improvement targets. © John Wiley & Sons, Inc. Naval Research Logistics 48: 684–709, 2001     

The military as a learning organisation: establishing the fundamentals of best-practice in lessons-learned

The post-Cold War era has witnessed the rapid expansion of organisational learning initiatives within NATO militaries, especially formal “lessons-learned” processes. The effectiveness of national lessons-learned processes in recalibrating military activity to the demands of ongoing operations has been highly-differentiated. However, the academic literature on military change and practitioner guidance has been slow to investigate the key features of best-practice in military learning. This article breaks new ground by drawing upon the literature on dynamic organisational capabilities to explore the fundamental organisational processes and activities which are necessary to implement successful lessons-learned. It examines, in particular, the organisational features which facilitate “knowledge transformation”: the effective combination of new knowledge gained from intra- and inter-organisational learning with existing organisational knowledge. The article concludes by highlighting several potential future empirical and theoretical research agendas in military learning and the importance of engagement between lessons-learned practitioners and organisational learning scholars.     

Call center arrival modeling: A Bayesian state‐space approach

In this article, we introduce three discrete time Bayesian state‐space models with Poisson measurements, each aiming to address different issues in call center arrival modeling. We present the properties of the models and develop their Bayesian inference. In so doing, we provide sequential updating and smoothing for call arrival rates and discuss how the models can be used for intra‐day, inter‐day, and inter‐week forecasts. We illustrate the implementation of the models by using actual arrival data from a US commercial bank's call center and provide forecasting comparisons. © 2011 Wiley Periodicals, Inc. Naval Research Logistics 58: 28–42, 2011     

Stochastic framework for partially degradation systems with continuous component degradation‐rate‐interactions

Many conventional models that characterize the reliability of multicomponent systems are developed on the premise that for a given system, the failures of its components are independent. Although this facilitates mathematical tractability, it may constitute a significant departure from what really takes place. In many real‐world applications, system components exhibit various degrees of interdependencies, which present significant challenges in predicting degradation performance and the remaining lifetimes of the individual components as well as the system at large. We focus on modeling the performance of interdependent components of networked systems that exhibit interactive degradation processes. Specifically, we focus on how the performance level of one component affects the degradation rates of other dependent components. This is achieved by using stochastic models to characterize how degradation‐based sensor signals associated with the components evolve over time. We consider “Continuous‐Type” component interactions that occur continuously over time. This type of degradation interaction exists in many applications, in which interdependencies occur on a continuum. We use a system of stochastic differential equations to capture such “Continuous‐Type” interaction. In addition, we utilize a Bayesian approach to update the proposed model using real‐time sensor signals observed in the field and provide more accurate estimation of component residual lifetimes. © 2014 Wiley Periodicals, Inc. Naval Research Logistics 61: 286–303, 2014     

Lot sizing with learning and forgetting in setups: Analytical results and insights

This article considers the dynamic lot sizing problem when there is learning and forgetting in setups. Learning in setups takes place with repetition when additional setups are made and forgetting takes place when there is a break between two successive setups. We allow the amount forgotten over a break to depend both on the length of the break and the amount of learning at the beginning of the break. The learning and forgetting functions we use are realistic. We present several analytical results and use these in developing computationally efficient algorithms for solving the problem. Some decision/forecast horizon results are also developed, and finally we present managerial insights based on our computational results. © 2016 Wiley Periodicals, Inc. Naval Research Logistics 63: 93–108, 2016     

基于稀疏贝叶斯学习的扩展目标雷达关联成像*

雷达关联成像不依赖于雷达与目标的相对运动,是一种高分辨凝视成像方式。传统的关联成像方法未考虑复杂扩展目标的结构信息,在高分辨成像时的应用受到限制,为此提出一种自适应结构配对稀疏贝叶斯学习方法。该算法在稀疏贝叶斯学习的框架内针对扩展目标建立一种结构配对层次化高斯先验模型,然后采用变分贝叶斯期望-最大化算法交替进行目标重构和参数优化。该算法将某一信号分量的重构与周围信号分量联系起来,并能在迭代过程中自适应地调整表征各信号分量相关性的参数。实验结果表明,该方法针对扩展目标可以有效地进行高分辨成像。     

基于贝叶斯决策的图像检索相关反馈方法研究

为了有效解决基于内容图像检索中“语义鸿沟”问题,利用贝叶斯决策理论估计图像检索中相关图像集与不相关图像集合的分类面,通过移动查询点的策略,指导新一轮的查询,将查询点移向好的范例点而远离坏的范例点。试验证明该方法是一种有效的基于统计分析的反馈方法。     

A new analysis of a lot-size model with partial backlogging

We reformulate the cost equation for the lot-size model with partial back-logging. The formulation is in terms of “fictitious demand rate,” a new inventory decision variable that simplifies the analysis. Using decomposition by projection, we obtain an optimal solution in a straightforward manner. The form of the solution sheds additional light on the behavior of the model. Some of these insights are elucidated by numerical examples.     

Bayesian inference for a Lanchester type combat model

We undertake inference for a stochastic form of the Lanchester combat model. In particular, given battle data, we assess the type of battle that occurred and whether or not it makes any difference to the number of casualties if an army is attacking or defending. Our approach is Bayesian and we use modern computational techniques to fit the model. We illustrate our method using data from the Ardennes campaign. We compare our results with previous analyses of these data by Bracken and Fricker. Our conclusions are somewhat different to those of Bracken. Where he suggests that a linear law is appropriate, we show that the logarithmic or linear‐logarithmic laws fit better. We note however that the basic Lanchester modeling assumptions do not hold for the Ardennes data. Using Fricker's modified data, we show that although his “super‐logarithmic” law fits best, the linear, linear‐logarithmic, and logarithmic laws cannot be ruled out. We suggest that Bayesian methods can be used to make inference for battles in progress. We point out a number of advantages: Prior information from experts or previous battles can be incorporated; predictions of future casualties are easily made; more complex models can be analysed using stochastic simulation techniques. © 2000 John Wiley & Sons, Inc. Naval Research Logistics 47: 541–558, 2000     

Bayesian strategies for dynamic pricing in e‐commerce

E‐commerce platforms afford retailers unprecedented visibility into customer purchase behavior and provide an environment in which prices can be updated quickly and cheaply in response to changing market conditions. This study investigates dynamic pricing strategies for maximizing revenue in an Internet retail channel by actively learning customers' demand response to price. A general methodology is proposed for dynamically pricing information goods, as well as other nonperishable products for which inventory levels are not an essential consideration in pricing. A Bayesian model of demand uncertainty involving the Dirichlet distribution or a mixture of such distributions as a prior captures a wide range of beliefs about customer demand. We provide both analytic formulas and efficient approximation methods for updating these prior distributions after sales data have been observed. We then investigate several strategies for sequential pricing based on index functions that consider both the potential revenue and the information value of selecting prices. These strategies require a manageable amount of computation, are robust to many types of prior misspecification, and yield high revenues compared to static pricing and passive learning approaches. © 2006 Wiley Periodicals, Inc. Naval Research Logistics, 2007     

Strategic capacity decision‐making in a stochastic manufacturing environment using real‐time approximate dynamic programming

In this study, we illustrate a real‐time approximate dynamic programming (RTADP) method for solving multistage capacity decision problems in a stochastic manufacturing environment, by using an exemplary three‐stage manufacturing system with recycle. The system is a moderate size queuing network, which experiences stochastic variations in demand and product yield. The dynamic capacity decision problem is formulated as a Markov decision process (MDP). The proposed RTADP method starts with a set of heuristics and learns a superior quality solution by interacting with the stochastic system via simulation. The curse‐of‐dimensionality associated with DP methods is alleviated by the adoption of several notions including “evolving set of relevant states,” for which the value function table is built and updated, “adaptive action set” for keeping track of attractive action candidates, and “nonparametric k nearest neighbor averager” for value function approximation. The performance of the learned solution is evaluated against (1) an “ideal” solution derived using a mixed integer programming (MIP) formulation, which assumes full knowledge of future realized values of the stochastic variables (2) a myopic heuristic solution, and (3) a sample path based rolling horizon MIP solution. The policy learned through the RTADP method turned out to be superior to polices of 2 and 3. © 2010 Wiley Periodicals, Inc. Naval Research Logistics 2010     

Neighborhood search heuristics for selecting hierarchically well‐formulated subsets in polynomial regression

The importance of subset selection in multiple regression has been recognized for more than 40 years and, not surprisingly, a variety of exact and heuristic procedures have been proposed for choosing subsets of variables. In the case of polynomial regression, the subset selection problem is complicated by two issues: (1) the substantial growth in the number of candidate predictors, and (2) the desire to obtain hierarchically well‐formulated subsets that facilitate proper interpretation of the regression parameter estimates. The first of these issues creates the need for heuristic methods that can provide solutions in reasonable computation time; whereas the second requires innovative neighborhood search approaches that accommodate the hierarchical constraints. We developed tabu search and variable neighborhood search heuristics for subset selection in polynomial regression. These heuristics are applied to a classic data set from the literature and, subsequently, evaluated in a simulation study using synthetic data sets. © 2009 Wiley Periodicals, Inc. Naval Research Logistics, 2010     

面向监督学习的稀疏平滑岭回归方法

岭回归是监督学习中的一个重要方法,被广泛用于多目标分类和识别。岭回归中一个重要的步骤是定义一个特殊的多变量标签矩阵,以实现对多类别样本的编码。通过将岭回归看作是一种基于图的监督学习方法,拓展了标签矩阵的构造方法。在岭回归的基础之上,进一步考虑投影中维度的平滑性和投影矩阵的稀疏性,提出稀疏平滑岭回归方法。对比一系列经典的监督线性分类算法,发现稀疏平滑岭回归在多个数据集上有着更好的表现。另外,实验表明新的标签矩阵构造方法不会降低原始岭回归方法的表现,同时还可以进一步提升稀疏平滑岭回归方法的性能。     

基于偏最小二乘回归的军用飞机采购价格预测

考虑到军用飞机采购价格样本数据少、难于预测的特点和偏最小二乘回归方法在处理小样本多元数据方面的优势,提出一种基于偏最小二乘回归的军用飞机价格预测方法.偏最小二乘回归首先提取第一、第二主成分对采购价格样本的特异点进行剔除;然后进行变量投影重要度分析来筛选变量;最后,偏最小二乘回归对筛选的变量进行回归建立军用飞机价格预测模型,并对军用飞机价格进行预测.结果表明,在军用飞机价格预测方面,与未筛选变量的回归模型和逐步多元回归相比,经过变量筛选的偏最小二乘回归模型预测的精度更高,更能体现采购价格与飞机性能参数之间的关系.     

Public opinion: A powerful predictor of U.S. defense spending

Public opinion survey responses regarding the desirability of changes in defense spending can be compressed into a single variable, the public opinion balance, which, when accompanied by a control variable measuring the proportion of responses in the “residuum” (no opinion or keep the status quo), permits an accurate prediction of subsequent changes in the rate of change of U.S. defense outlays from the mid‐1960s through the 1980s. This finding cannot be interpreted as a simple case of “the public got what it wanted,” however, because public opinion was not autonomous or spontaneous, and defense decision makers themselves played a central role in shaping public opinion.     

Approximating the nondominated frontiers of multi‐objective combinatorial optimization problems

Finding all nondominated vectors for multi‐objective combinatorial optimization (MOCO) problems is computationally very hard in general. We approximate the nondominated frontiers of MOCO problems by fitting smooth hypersurfaces. For a given problem, we fit the hypersurface using a single nondominated reference vector. We experiment with different types of MOCO problems and demonstrate that in all cases the fitted hypersurfaces approximate all nondominated vectors well. We discuss that such an approximation is useful to find the neighborhood of preferred regions of the nondominated vectors with very little computational effort. Further computational effort can then be spent in the identified region to find the actual nondominated vectors the decision maker will prefer. © 2009 Wiley Periodicals, Inc. Naval Research Logistics, 2009     

Detection of minimal forecast horizons in dynamic programs with multiple indicators of the future

Many sequential planning problems can be represented as a shortest path problem in an acyclic network. This includes all deterministic dynamic programs as well as certain stochastic sequential decision problems. In this article, we identify a large class of shortest path problems for which a general efficient algorithm for the simultaneous solution and detection of minimal forecast horizons is developed. Detection of a such minimal forecast horizons is essential when accurate information regarding various relevant parameters is obtained progressively, i.e., when the initial information is restricted to a limited horizon of “future” stages only. We describe five classes of planning problems which can be efficiently addressed by the general algorithm. These classes deal with multi-item joint replenishment systems, combined inventory and routing problems, machine scheduling issues, single item stochastic inventory settings and routing problems in the plane and in space. © 1996 John Wiley & Sons, Inc.     

THE CONCEPT OF NUCLEAR LEARNING

This article seeks to elucidate the concept of nuclear learning. It explores both the “nuclear” and the “learning” aspects of the concept. On the nuclear side, it distinguishes between learning basic facts about nuclear arms and drawing inferences about the larger implications of those facts. On the learning side, it discusses three issues: whether to use the term in a normative or value-neutral manner; the difference between learning that leads to a change in means versus learning that leads to a re-evaluation of ends; and whether learning only takes place at the level of individuals or whether there can also be learning by collective entities. The article argues there is no universal best answer to these questions and that the particular concept of learning that should be employed depends on the goals of the analyst. If the goal is to reduce the chances of nuclear war, however, one type of learning that will be important to consider is whether there is shared, cross-national learning.     

Concept costing for defence projects: the problem and its solution

It is intrinsic to the nature of military competition that the unit acquisition costs of defence systems rise generation by generation. Traditional “bottom‐up” methods of cost‐estimating cannot meet the increasing demands for extensive studies of many options before commencing design and, while “top‐down” estimating techniques are an improvement, in particular by reducing the time required for preparing estimates, they also are inadequate to meet current requirements. A new method has been devised, therefore, further developing “top‐down” methods using Bayesian techniques to make best use of the available information, whether certain or uncertain, and its accuracy established by example.