一种用于知识发现的非精确学习算法

海量数据的快速增长,伴随着数据挖掘技术在人工智能,机器学习与知识工程等领域进入了深层次的研究阶段,从而为非结构数据的知识发现提供了可能。本文采用形式化方法详细地描述了关于非精确规则获取的基本原理和知识发现过程,并提出了一种应用于值为数据类型的知识发现的非精确学习模型,实验表明该算法在非结构数据知识发现领域具有一定的可行性与实用性。

基于梯度下降的脉冲神经元精确序列学习算法

基于梯度下降的脉冲神经元有监督学习算法通过计算梯度最小化目标序列和实际输出序列间的误差使得神经元能激发出目标脉冲序列。然而该算法中的误差函数是基于实际输出脉冲序列和相对应的目标输出脉冲序列动态构建而成,导致算法在收敛时可能出现实际输出序列的个数和期望输出个数不相等的情况。针对这一缺陷提出了一种改进的脉冲神经元梯度下降学习算法,算法在学习过程中检测目标序列脉冲个数和实际激发脉冲个数,并引入虚拟实际激发脉冲和期望激发脉冲构建误差函数以分别解决激发个数不足和激发个数多余的问题。实验结果证明该算法能有效地防止学习算法在输出脉冲个数不等的情况下提前结束,使得神经元能够精确地激发出目标脉冲序列。

The Habitus Made Me Do It： Bourdieu＇s Key Concept as a Substruction of the Monad

My claim is that Bourdieu＇s concept of habitus is not consistent and its ambiguities conceal an imprecision concerning the subject of social action. Indeed, Bourdieu defines habitus in three different ways： as a capacity, as a set of dispositions, and as a scheme for practice. That is why he cannot solve the problem of the duality of agent and habitus and produces a problem offundamentation, as we can see in his troubles to determine which is the substratum of social actions. Though Bourdieu claims he borrows the concept of habitus from Husserl and other phenomenologists, many divergences can be stated in the way they conceive it. Unlike Bourdieu, phenomenology can establish precise relations of fundamentation between agent, habitus, and the ego because it avoids the fallacy of the wrong level involved in the attribution of systemic properties to personal eogic structures. Accordingly, it provides a consistent and precise concept of the habitus,