The learning rate of l 2-coefficient regularized classification with strong loss

In the present paper, we give an investigation on the learning rate of l2-coefficient regularized classification with strong loss and the data dependent kernel functional spaces. The results show that the learning rate is influenced by the strong convexity.     

Pi-sigma神经网络的带动量项的异步批处理梯度算法收敛性

本文将动量项引入到训练Pi-sigma神经网络的异步批处理的梯度算法中,有效的改善了算法的收敛效率,并从理论上对该算法的收敛性进行研究,给出了误差函数的单调性定理及该算法的弱收敛和强收敛性定理.计算机仿真实验亦验证了带动量项的异步批处理梯度算法的有效性和理论分析的正确性.     

Rejoinder on “Imprecise probability models for learning multinomial distributions from data. Applications to learning credal networks”

In this paper we answer to the comments provided by Fabio Cozman, Marco Zaffalon, Giorgio Corani, and Didier Dubois on our paper ‘Imprecise Probability Models for Learning Multinomial Distributions from Data. Applications to Learning Credal Networks’. The main topics we have considered are: regularity, the learning principle, the trade-off between prior imprecision and learning, strong symmetry, and the properties of ISSDM for learning graphical conditional independence models.     

Considering Ethnicity in Teaching and Learning Statistics: Should I Worry About Where my Students Come From?

In the past decade there has been strong interest in the special needs of overseas students attending Australian universities, with respect to teaching and learning. This paper reports on three action research studies that address the question of whether such issues remain in the teaching and learning of statistics in particular.     

多层感知器信用评模型及预警研究

本文利用多层感知器 ( MLP)原理建立神经网络信用评价模型 ,用来对我国 2 0 0 0年 1 0 6家上市公司进行信用评级 ,并进一步对我国 2 0 0 1年公布的 1 3家预亏公司进行预警研究 .按照各上市公司的经营状况分为“好”、“差”两类 ,每一类由 5 3家上市公司构成数据样本 .对于每一家上市公司 ,主要考虑其经营状况的四个财务指标 :每股收益 ,每股净资产 ,净资产收益率和每股现金流量 .仿真结果表明 ,本文所建立的神经网络信用评价模型有很高的分类准确率 ,达到 98.1 1 % .又由于该信用评价模型有很强的适应能力 ,故可以进一步用来对企业的财务危机进行预警研究 .预警实证分析表明 ,该信用评价模型对我国 2 0 0 1年公布的 1 3家预亏公司进行预警分析 ,预警准确率达到 1 0 0 % .此外 ,文中还给出 MLP网络模型的学习算法和步骤     

Learning intransitive reciprocal relations with kernel methods

In different fields like decision making, psychology, game theory and biology, it has been observed that paired-comparison data like preference relations defined by humans and animals can be intransitive. Intransitive relations cannot be modeled with existing machine learning methods like ranking models, because these models exhibit strong transitivity properties. More specifically, in a stochastic context, where often the reciprocity property characterizes probabilistic relations such as choice probabilities, it has been formally shown that ranking models always satisfy the well-known strong stochastic transitivity property. Given this limitation of ranking models, we present a new kernel function that together with the regularized least-squares algorithm is capable of inferring intransitive reciprocal relations in problems where transitivity violations cannot be considered as noise. In this approach it is the kernel function that defines the transition from learning transitive to learning intransitive relations, and the Kronecker-product is introduced for representing the latter type of relations. In addition, we empirically demonstrate on two benchmark problems, one in game theory and one in theoretical biology, that our algorithm outperforms methods not capable of learning intransitive reciprocal relations.     

基于一阶Takagi-Sugeno系统的Pi-Sigma网络的改进神经-模糊梯度学习算法

This paper presents a Pi-Sigma network to identify first-order Tagaki-Sugeno（T-S） fuzzy inference system and proposes a simplified gradient-based neuro-fuzzy learning algorithm.A comprehensive study on the weak and strong convergence for the learning method is made,which indicates that the sequence of error function goes to a fixed value,and the gradient of the error function goes to zero,respectively.     

A depletion of assets model of organizational learning*

This research investigates how population heterogeneity with respect to initial assets affects the rate of failure (or change) of organizations and social relationships. Organizations and social relationships are assumed to be endowed with initial assets which buffer against initial risks of failure. Failure is seen as the outcome of a process in which the assets are depleted and finally become exhausted. The core idea is that the assets of a unit become depleted through learning experiences which incur search costs and setbacks. Two settings are explored: A constant rate of depletion, and a declining rate of depletion of assets. The study explores additionally how the distribution of initial assets affects the failure rate of the population. It is found that the type of distribution of initial assets has a strong impact on the time dependence of the failure rate. A Normal distribution of initial assets leads to positive time dependence if the depletion rate is constant, and to negative time dependence if the depletion rate declines. The results show that population heterogeneity with respect to initial assets has effects on the time dependence of failure rates which are quite different from the popular case of population heterogeneity with respect to fit.     

Regularized least square regression with dependent samples

In this paper we study the learning performance of regularized least square regression with α-mixing and ϕ-mixing inputs. The capacity independent error bounds and learning rates are derived by means of an integral operator technique. Even for independent samples our learning rates improve those in the literature. The results are sharp in the sense that when the mixing conditions are strong enough the rates are shown to be close to or the same as those for learning with independent samples. They also reveal interesting phenomena of learning with dependent samples: (i) dependent samples contain less information and lead to worse error bounds than independent samples; (ii) the influence of the dependence between samples to the learning process decreases as the smoothness of the target function increases.     

基于学习向量量化神经网络的水环境质量综合评价

针对水环境质量综合评价中指标权重确定方法的不足,利用学习向量量化(LVQ)神经网络具有的强大的非线性运算和相似特征聚类功能,提出一种基于学习向量量化(LVQ)神经网络的水质综合评价决策方法.将它应用于水质综合指标评价,为改进水质综合评价提供了一种简捷的分类评价方法.     

Single Machine Scheduling with Learning Effect Considerations

In this paper we study a single machine scheduling problem in which the job processing times will decrease as a result of learning. A volume-dependent piecewise linear processing time function is used to model the learning effects. The objective is to minimize the maximum lateness. We first show that the problem is NP-hard in the strong sense and then identify two special cases which are polynomially solvable. We also propose two heuristics and analyse their worst-case performance.     

人工神经网络BP算法的改进和结构的自调整

本文解决了BP神经网络结构参数和学习速率的选取问题,并对传统的BP算法进行了改进,提出了BP神经网络动态全参数自调整学习算法,又将其编制成计算机程序,使得隐层节点和学习速率的选取全部动态实现,减少了人为因素的干预,改善了学习速率和网络的适应能力。计算结果表明：BP神经网络动态全参数自调整算法较传统的方法优越。训练后的神经网络模型不仅能准确地拟合训练值,而且能较精确地预测未来趋势。     

Common due date scheduling with autonomous and induced learning

Up to now the few existing models, that consider learning effects in scheduling, concentrate on learning-by-doing (autonomous learning). But recent contributions to the literature on learning in manufacturing organizations emphasize the important impact of proactive investments in technological knowledge on the learning rate (induced learning). In the present paper, we focus on a scheduling problem where the processing times decrease according to a learning rate, which can be influenced by an initial cost-inducing investment. Thus we have integrated into our model both aspects of learning––autonomous and induced––thereby highlighting the management's responsibility to invest in technological knowledge enhancement. We have been able to derive some structural properties of the problem and present a polynomially bound solution procedure which optimally solves the problem by using these properties. The optimal solution to the scheduling problem contains––of course–– information on the optimal level of proactive investments in learning.     

Accelerating autonomous learning by using heuristic selection of actions

This paper investigates how to make improved action selection for online policy learning in robotic scenarios using reinforcement learning (RL) algorithms. Since finding control policies using any RL algorithm can be very time consuming, we propose to combine RL algorithms with heuristic functions for selecting promising actions during the learning process. With this aim, we investigate the use of heuristics for increasing the rate of convergence of RL algorithms and contribute with a new learning algorithm, Heuristically Accelerated Q-learning (HAQL), which incorporates heuristics for action selection to the Q-Learning algorithm. Experimental results on robot navigation show that the use of even very simple heuristic functions results in significant performance enhancement of the learning rate.     

Spaces to discuss mathematics: communities of practice on an online discussion board

In theories of learning that adopt a situated stance to knowledge the notion of identity is vital; how learners position themselves in relation to, and are mutually positioned by, the situation within which they are learning will have a strong bearing on the learning outcomes. One of the challenges for learning mathematics in school is that learners position themselves, and are positioned, as pupils rather than as mathematicians. This paper focuses on discussion boards designed for secondary school mathematics students, and we use Wenger's (1998) model of communities of practice, building on earlier work by the authors (Back and Pratt 2007; Pratt and Kelly 2007) in which ‘idealised communities’ are constructed and used, to consider a case study of one participant who engages in developing his identity as a mathematician doing mathematics, as well his identity as a learner and a teacher of mathematics.     

Learning angles through movement: Critical actions for developing understanding in an embodied activity

Angle instruction often begins with familiar, real-world examples of angles, but the transition to more abstract ideas can be challenging. In this study, we examine 20 third and fourth grade students completing a body-based angle task in a motion-controlled learning environment using the Kinect for Windows. We present overall pre- and post-test results, showing that the task enhanced learners’ developing ideas about angles, and we describe two case studies of individual students, looking in detail at the role the body plays in the learning process. We found that the development of a strong connection between the body and the abstract representation of angle was instrumental to learning, as was exploring the space and making connections to personal experiences. The implications of these findings for developing body-based tasks are discussed.     

状态转移下的利率预期与误差学习假说

采用线性回归和状态转移模型讨论利率预期的"误差学习假说"和不同期限利率预测误差之间的关联性,并考虑带有风险溢价时预测误差的作用.分析发现,1年期限的"误差学习假说"在银行间国债市场中是成立的,不同期限的利率预期对预测误差的反应呈现一定的独立性,不同的状态中预测误差和风险溢价对利率预期的影响是不同的,考虑风险溢价有助于提升对利率预期变动的解释.     

Mathematical tasks,study approaches,and course grades in undergraduate mathematics: a year-by-year analysis

Students approach learning in different ways, depending on the experienced learning situation. A deep approach is geared toward long-term retention and conceptual change while a surface approach focuses on quickly acquiring knowledge for immediate use. These approaches ultimately affect the students’ academic outcomes. This study takes a cross-sectional look at the approaches to learning used by students from courses across all four years of undergraduate mathematics and analyses how these relate to the students’ grades. We find that deep learning correlates with grade in the first year and not in the upper years. Surficial learning has no correlation with grades in the first year and a strong negative correlation with grades in the upper years. Using Bloom's taxonomy, we argue that the nature of the tasks given to students is fundamentally different in lower and upper year courses. We find that first-year courses emphasize tasks that require only low-level cognitive processes. Upper year courses require higher level processes but, surprisingly, have a simultaneous greater emphasis on recall and understanding. These observations explain the differences in correlations between approaches to learning and course grades. We conclude with some concerns about the disconnect between first year and upper year mathematics courses and the effect this may have on students.     

一类供应链中考虑公平关切的学习效应行为实验研究

大量经济学实验研究证实了公平关切和学习效应对决策者行为的影响力。本文研究三人组供应链系统,通过区别设计个体自我学习以及社会学习的实验环境,对比考察备用供应商的公平关切程度,以及制造商和备用供应商学习曲线的特点。实验结果支持了学习效应存在的假设:随着实验期数的增加,单期决策时间逐渐减少,备用供应商的整体拒绝率逐渐降低,制造商的策略逐渐集中。进一步构建了引入公平关切的强化学习模型。通过参数估计发现在个体自我学习和社会学习实验环境下,备用供应商的横向公平关切程度均较为显著,信息共享对备用供应商的横向公平关切偏好无明显影响。