Symbol Processing System Modeled after Brains
Symbol Processing System Modeled after Brains
批准号:
15500095
负责人:
SAKURAI Akito
金额:
$2.37万
依托单位:
依托单位国家:
日本
项目类别:
Grant-in-Aid for Scientific Research (C)
财政年份:
2003
资助国家:
日本
项目状态:
已结题
起止时间:
2003 至 2005
中文摘要
我们研究了一种新的神经网络学习框架,该框架由多个强化学习代理组成,其中存在多个合法的候选模块。我们发明了一种促进强化学习智能体之间竞争学习的机制,并通过计算机模拟验证了它的有效性。我们进一步研究了另一种类型的语法习得,即两种不同类型的递归神经网络。它们中的一种类型监视另一种类型,并根据监视的观测进行自我修改。我们以宾夕法尼亚树库的ATIS语料库为语料库,以ILP为基本学习范式,进行了移位-归约句法器的习得实验。为了缓解现有学习方法的缺点(执行时间和内存要求较高),我们使用语法类别作为学习单元。我们发明了新的方法来根据语法范畴生成合理化的否定例子,并通过调查这些例子实际上被遗漏分类的地方来重新学习这些否定例子。我们证实,有限状态自动机的准确率提高到了略低于90%的水平,能够在大脑中表示知识,但众所周知,它们的通用性太强,无法成功地学习。因此,我们研究了近似地学习和交流它们的方法。我们应用了强化学习方法,发现它们是合格的。我们发明了一种方法来准备大量的语法类别,试图在交际中使用它们,并选择最好的类别。在递归神经网络中实现了该方法,并进行了数值模拟。从只关注训练错误(而不是泛化能力)的意义上重构了原有的语法范畴结构,结果是令人振奋的。
英文摘要
We have investigated a new framework of neural network learning that is composed of multiple reinforcement learning agents among which there exist multiple legitimate candidate modules. We have invented a mechanism that facilitates competitive learning among reinforcement learning agents and ascertained its validity by computer simulations.We further investigated another type of grammar acquisition by recurrent neural networks of two different types. One type of them monitors the other type and modifies itself based on the monitored observation. We found that the networks are able to learn grammatical categories and are robust against their legion.We conducted experiments of acquisition of shift-reduce parsers in which ATIS corpus in Penn TreeBank is the corpus and ILP is the fundamental learning paradigm. To alleviate drawbacks (high cost of execution time and memory requirements) of the existing learning methods, we employed grammatical categories as learning units. We invented new methods to generate rationalized negative examples based on grammatical categories and to relearn the negative examples by investigating where those examples are in fact miss-classified. We confirmed that the accuracy improved to a bit less than 90%,Finite state automata are capable enough to represent knowledge in brain but it is well-known that they are too versatile to be successfully learned. Therefore we made research on methods to approximately learn and communicate them. We have applied reinforcement learning methods and found them to be eligible. We have invented a method to prepare a large number of grammatical categories, to try to use them in communication, and to select best ones. We implemented the method in recurrent neural networks, and conducted numerical simulations. The results are promising in a sense that the original grammatical category structure is reconstructed with paying attention only to training errors (not to generalization capabilities).
期刊论文(22)
专著(0)
科研奖励(0)
会议论文
登录
查看更多内容
DOI:
--
发表时间:
2005
期刊:
Proc. Second Int. Symp. Emergence and Evolution of Linguistic Communication 1
影响因子:
--
作者:
[Y.Shinozawa, A.Sakurai]
通讯作者:
A.Sakurai
Frame Net-Based Shallow Semantic Parsing with a POS Tagger
使用词性标注器构建基于网络的浅层语义解析
DOI:
--
发表时间:
2004
期刊:
Proc. Joint Workshop of Vietnamese Society of AI, SIGKBS-JSAI, ICS-IPSJ and IEICE-SIGAI on Active Mining 1
影响因子:
--
作者:
[N.Shibui, A.Sakurai]
通讯作者:
A.Sakurai
FrameNet-Based Shallow Semantic Parsing with a POS Tagger
使用 POS 标记器进行基于 FrameNet 的浅层语义解析
DOI:
--
发表时间:
2005
期刊:
Proc.Joint Workshop of Vietnamese Society of AI, SIGKBS-JSAI, ICS-IPSJ and IEICE-SIGAI on Active Mining
影响因子:
--
作者:
[N.Shibui, A.Sakurai]
通讯作者:
A.Sakurai
A Role Sharing Model of Language Areas
语言区域的角色共享模型
DOI:
--
发表时间:
2006
期刊:
Proceedings of First International Workshop on Emergence and Evolution of Linguistic Communication 1(未定)
影响因子:
--
作者:
[Y.Shinozawa, A.Sakurai]
通讯作者:
A.Sakurai
強化学習におけるオンラインセンサ選択
强化学习中的在线传感器选择
DOI:
--
发表时间:
2005
期刊:
電気学会論文誌C 125-6
影响因子:
--
作者:
[石川 浩一郎, 櫻井 彰人, 藤波 努, 國藤 進]
通讯作者:
國藤 進
共 9 条
A proposal of structural mixture distribution model. its application and basic analysis-
-
批准号:21500146
-
项目类别:Grant-in-Aid for Scientific Research (C)
-
资助金额:$2.83万
-
财政年份:2009
-
负责人:SAKURAI Akito
-
依托单位:
Time series information processing by networking finite state neural networks
-
批准号:18500118
-
项目类别:Grant-in-Aid for Scientific Research (C)
-
资助金额:$2.6万
-
财政年份:2006
-
负责人:SAKURAI Akito
-
依托单位:
Symbol Processing System Modeled after Brains
-
批准号:13680438
-
项目类别:Grant-in-Aid for Scientific Research (C)
-
资助金额:$1.98万
-
财政年份:2001
-
负责人:SAKURAI Akito
-
依托单位:
海外基金