混合模糊语义细胞的学习及其应用

发布时间：2018-04-29 02:44

本文选题：概念 + 模糊语义细胞　；参考：《浙江大学》2017年硕士论文

【摘要】：概念实体的表达往往具有一定的模糊性,这种模糊性是蕴含在在语义中出现的。使用合适的概念模型来表达模糊语义具有重要的意义。模糊语义细胞作为最小的模糊概念的表示单元,在数据挖掘、机器学习以及知识发现中具有重要的作用。在概念空间(论域)Ω上,模糊语义细胞L =P,d,δ被称为"关于pi","类似Pi"以及"和Pi接近"的语义标签,其中P代表概念i的原型,d是定义在论域Ω上的距离函数,δ则是概念空间中定义在[0,+∞)上其他点和Pi的距离的概率密度函数。在模糊语义细胞的学习中我们需要关注语义的覆盖程度、描述的清晰程度以及描述的模糊性这三个因素,因此模糊语义细胞的学习原则很自然地就联系到最大覆盖率、最具典型性和最大模糊熵这三个指标之上。本文中混合模糊语义细胞是建立在模糊语义细胞的学习基础之上,模糊语义细胞学习的最终目标是要寻找最佳的L来刻画具有某个概念的数据集,而混合模糊语义细胞则在此基础上做了更深一层的拓展,考虑具有若干个相关的概念的集合LA={L1,L2,...,Ln},其中每个概念都对应使用模糊语义细胞Li来描述第i个概念的数据集,混合模糊语义细胞的学习是为了能够寻找到一组最合适的权重参数W={W1,w2,...w2}来刻画某个概念在此概念集合(主题)中的影响程度或者是重要程度,借鉴之前的模糊语义细胞的学习原则,我们需要重新定义并计算语义细胞的两个数字特征:期望粒度R和模糊熵H。最终将学习混合模糊语义细胞的问题转化为了非线性约束优化问题。
[Abstract]:The expression of conceptual entities often has some fuzziness, which is contained in semantics. It is of great significance to express fuzzy semantics with appropriate conceptual models. As the smallest representation unit of fuzzy concepts, fuzzy semantic cells play an important role in data mining, machine learning and knowledge discovery. On the concept space (domain) 惟, the fuzzy semantic cell L Pu D, 未 is called the semantic label "on pi", "similar to Pi" and "close to Pi". Where P represents the prototype of the concept I / d is the distance function defined on the domain 惟, and 未 is the probability density function of the distance between other points and Pi defined on [0, 鈭瀅 in the concept space. In the learning of fuzzy semantic cells, we need to pay attention to the three factors of semantic coverage, clarity of description and fuzziness of description, so the learning principle of fuzzy semantic cells is naturally related to the maximum coverage. The most typical and the maximum fuzzy entropy above these three indicators. In this paper, mixed fuzzy semantic cells are based on the learning of fuzzy semantic cells. The ultimate goal of fuzzy semantic cell learning is to find the best L to depict the data set with a certain concept. On the other hand, the mixed fuzzy semantic cells are further extended to consider the set of several related concepts LA= {L1 / L2U. N}, in which each concept corresponds to the data set in which the fuzzy semantic cell Li is used to describe the first concept. The learning of mixed fuzzy semantic cells is to be able to find the most appropriate set of weight parameters W = {W1W2U. W2} to describe the degree of influence or importance of a concept in this set of concepts (themes). Based on the learning principle of fuzzy semantic cells, we need to redefine and calculate the two numerical features of semantic cells: expected granularity R and fuzzy entropy H. Finally, the problem of learning mixed fuzzy semantic cells is transformed into nonlinear constrained optimization problem.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1

【参考文献】