63 research outputs found
Narratives of Furui 符瑞 and the Compilation of Dynastic Histories in the Han and the Jin Periods: Focus on The Zhi 志 Framework of Narratives
No full textFurui (符瑞), auspicious phenomena, were recounted within the framework of Fengshan (封禅) sacrifices in the Shiji. This form of recounting furui was continued in the Hanshu within the framework of the treatise on the Rites of Suburban Sacrifice (郊祀). After the annals of the Xuan Di (宣帝) period in the Hanshu, it became standard practice to recount a series of measures in response to the holding of suburban rituals and the emergence of furui. On the other hand, as the Rites of Suburban Sacrifice became institutionalized in the late period of the Former Han, the role of furui changed, as evidenced by the Baihu Guan Council (白虎觀會議) held during the Later Han Emperor Zhangdi's reign. At this meeting, thirty-seven events were selected as furui to be recognized by the dynasty. In addition, their appearance was seen as being in response to the emperor's virtuous rule. This can be traced not only in the narrative of the Dongguan Hanji (東觀漢記), but also in the inscriptions that were popular at the time. The narratives of furui, which had extended to local communities, once again converged in the court from the end of the Later Han Dynasty to the Three Kingdoms period. In particular, furui that appeared at the time of the regime change between the Han and Wei dynasties are described as being closely related to the chengwei (讖雲). However, in the Xu Hanshu (續漢書) and Weishu (魏書) of the Sanguo zhi (三國志), which were compiled during the Western Jin dynasty, furui are described in a very restrained manner. This narrative trend is prominent in the Xu Hanshu. In addition, the Xu Hanshu not only does not recount furui but also describes the events that had been described as auspicious omens in Dongguan Hanji as disasters. Although it is difficult to say whether the attitude of omitting furui in the Xu Hanshu and Sanguo zhi was based on the judgment of the dynasty or the author himself, but it can be said to represent the political and social climate of the period. On the other hand, furui continued to be a subject of commentary in court diaries (起居注) of the Western Jin dynasty and in records related to the dynasty. This means that furui described in the Xu Hanshu and Sanguo zhi do not reflect the historical characteristics of the period they were written about
The portrait of the "hero" in the novels of Furui Yoshikichi and Furuyama Komao.
Full text linkpdfThe speaker focuses on the literature of the 1970s and the so-called inward-looking generation which represents that decade. With reference to characters of novels from this period he compiles a portrait of the character-type which epitomizes this generation.
(1) The Furui Yoshikichi's hero:
―living in an "extraordinary mundanity. "
The hero's sense of time and space
―the newly settled frontier region as a special kind of space.
―the confluence of traditional and modern cultures.
Relations between the sexes
―the highly-strung, abnormal woman, the man lacking in ambition.
(2) Furuyama Komao's diminutive protagonist
―the ordinariness of everyday life and the illusion of the past in Jimen o hau ningen (The human who crawls on the ground).
My research on the works of Furui and Furuyama was motivated by my interest in "the hero" in Japanese literature. This is an interest I have been pursuing for the past ten years, and about which I have wrritten several theses, for example, "The hero in the pōetry of Hagiwara Sakutaro", "The hero in the novels of Tanizaki Junichirō," and "Man as portrayed in the novels of Kaikō Ken."
From readings of Various works by a particular author, one gains an impression of an over-all character type (which may differ to some extent from novel to novel). The specific characteristics of various characters can be collated and, from these, qualities of human character can be extracted, so that various individual characters are reduced to a composite, abstract type. This procedure may be queried, but as time today is limited, I will present only a summary of my thesis, and only the results of my analyses will be explained in any detail. In other words, my remarks will be limited to an explication of what I consider to be the most salient aspects of the Furui Yoshikichi and Furuyama Komao herose.conference pape
招待発表（2） 古井由吉・古山高麗雄の小説の主人公
No full textpdfThe speaker focuses on the literature of the 1970s and the so-called inward-looking generation which represents that decade. With reference to characters of novels from this period he compiles a portrait of the character-type which epitomizes this generation.
(1) The Furui Yoshikichi's hero:
―living in an "extraordinary mundanity. "
The hero's sense of time and space
―the newly settled frontier region as a special kind of space.
―the confluence of traditional and modern cultures.
Relations between the sexes
―the highly-strung, abnormal woman, the man lacking in ambition.
(2) Furuyama Komao's diminutive protagonist
―the ordinariness of everyday life and the illusion of the past in Jimen o hau ningen (The human who crawls on the ground).
My research on the works of Furui and Furuyama was motivated by my interest in "the hero" in Japanese literature. This is an interest I have been pursuing for the past ten years, and about which I have wrritten several theses, for example, "The hero in the pōetry of Hagiwara Sakutaro", "The hero in the novels of Tanizaki Junichirō," and "Man as portrayed in the novels of Kaikō Ken."
From readings of Various works by a particular author, one gains an impression of an over-all character type (which may differ to some extent from novel to novel). The specific characteristics of various characters can be collated and, from these, qualities of human character can be extracted, so that various individual characters are reduced to a composite, abstract type. This procedure may be queried, but as time today is limited, I will present only a summary of my thesis, and only the results of my analyses will be explained in any detail. In other words, my remarks will be limited to an explication of what I consider to be the most salient aspects of the Furui Yoshikichi and Furuyama Komao herose
Plasmonic Black Absorbers for Enhanced Photocurrent of Visible-Light Photocatalysis
No full text
Plasmonic resonance of noble metal nanoparticles can drastically enhance the visible response of wide-bandgap photocatalysts such as TiO2, but the current technology has two fundamental problems: narrow absorption band and low absorption, which limit the energy efficiency of photocatalysis using sunlight. Here, an original plasmonic black absorber is reported, which sandwiches a 150 nm TiO2 layer between a layer of random Au nanoparticles and a rough Au surface (200 nm thick). The combined plasmonic effect of the Au nanoparticles and the Au rough surface enables a strong absorption (72%-91%) over 400-900 nm and a significantly (20-fold) enhanced photocurrent as compared to the bare TiO2 film. The strong absorption to visible and near infrared light, and the much enhanced photocurrent make the black absorber an ideal material for solar applications such as photocatalytic, photosynthetic, photovoltaic, and photothermal systems.</p
Robust Optimization Considering Probabilistic Magnetic Degradation
Full text linkThis paper presents robust topology optimization of electromagnetic machines by considering the magnetic degradation caused by mechanical and thermal stresses in punching, shrinking fitting, and other manufacturing processes. The topology optimization is performed using two methods: one is the robust genetic algorithm in which random noises are added to the magnetic characteristic parameters and the other takes the deviations in the objective and constraint functions due to the degradation into account. These methods are applied to optimization of the flux barrier shapes in an interior permanent magnetic motor to find that one can successfully realize robust design
Ji yu guan ce lian he fen bu de yin guo tui duan yu yin gong yin zi jian ce
No full textPh.D.Given the observational data of two causally related variables, distinguishing the cause variable from the effect variable is an emerging research area. Targeting different data types (discrete, one dimensional continuous or high dimensional continuous), a bunch of functional and nonfunctional models have been proposed. They infer the causal direction using so called cause-effect asymmetry, that is, statistical asymmetrical behaviors of the joint distribution when one examines it in causal versus anticausal directions. The asymmetry has different statistical definitions in different models. Very recently, researchers also extend to the situation that there is a confounder between cause and effect, and detect confounding based on the asymmetric behavior of the regression vector induced spectral measure in causal and confounding cases, when certain model assumptions are satisfied. In this thesis, we follow the asymmetry based inference framework, and propose several novel models. We propose two causal models for discrete data: one for ordered discrete data and the other for categorical data. For high dimensional continuous data, we propose a kernel based causal model. Finally we provide improvements to the confounding model.For ordered discrete data, we generalize the additive noise model to a mixture model, allowing multiple functions to exist. This model can deal with the situation that data is assembled from multiple heterogeneous environments, and in each environment the effect is a function of the cause plus a noise. To infer the causal direction, we rely on a novel statistical asymmetry that is defined in terms of support of the joint distribution. We show that the numbers of support points for the conditional distributions p(effect|cause) are with smaller variations compared to those of p(cause|effect). It is proved under the assumption that the minimum number of support points for p(cause|effect) is very small, justified by the fact that functions often have very few support points for their minimum and maximum values. The effectivity of the proposed method is also showed by empirical results.Another type of discrete data is categorical data, which is without order information. By shifting from functional model to nonfunctional model, we develop an efficient method to cover this scenario, using a novel statistical asymmetry based on direct independence tests. We consider p(cause) and p(effect|cause) as independent random variables taking realizations at each possible status of the cause, and use distance correlation to measure their dependence. Because p(cause) is independent with p(effect|cause) while p(effect) is dependent with p(cause|effect), distance correlation is smaller in the causal direction. Experiments and computational complexity analysis are conducted to show the effectiveness and efficiency of this inference method.In the era of big data, people need to conduct causal inference on high dimensional variables. Different from aforementioned scenarios, statistical asymmetry here should be defined in a way such that estimating the joint distribution is avoided. Thus, we propose to use kernel embedding, and give a novel statistical asymmetry based on freeness condition. The freeness condition is that the covariance matrix of the mean embedding of the cause in reproducing kernel Hilbert space is free independent with the covariance matrix of the conditional embedding of the effect given cause. This induces a cause-effect asymmetry that a designed measurement is 0 in the causal direction but smaller than 0 in the anticausal direction. We thoroughly study the identifiability of the causal direction using the proposed method, and show that the non-identifiable condition is rarely satisfied. Extensive empirical results provide evidence for our theoretical claims.Lastly, we shift from cause-effect asymmetry to causal-confounding asymmetry to detect a confounder between cause and effect. Recently proposed linear confounding model assumes that the model parameters are drawn uniformly at random, and identifies the existence of a confounder using the asymmetric behavior of regression vector induced spectral measures in causal and confounding cases. It is uniform in purely causal cases but non-uniform in the presence of a scalar confounder. Instead of analyzing the pattern of the spectral measure, we analyze its first moment. We compare its first moment with the first moment of a uniform reference spectral measure. The two moments coincide in causal cases, and differ from each other in the presence of confounding. This statistical causal-confounding asymmetry can be used for confounder detection. Our method is efficient and can better work in practical cases where spectral measure is often not ideally uniform. Experiments on both synthetic and real data are conducted to show the effectivity and efficiency of the proposed method.給定兩個因果相關變量的觀測數據，將因變量與果變量區分開來是一個新興的研究領域。針對不同的數據類型(離散的，一維連續的或高維連續的)，研究人員已經提出了一些函數化和非函數化的模型。他們使用所謂的因果不對稱性來推斷因果方向，即當人們從因果方向與反因果方向檢查聯合分佈時，它具有不對稱的統計性質。不對稱性質在不同模型中具有不同的統計定義。最近，研究人員還擴展到因和果之間存在隱共因子的情況，並基於無隱共因子和存在隱共因子情況下，給定模型假設成立下的回歸矢量產生的譜測度的不對稱性來檢測隱共因子。在本博士論文中，我們基於不對稱性的推斷框架，提出了幾個新的模型。我們提出了兩種離散數據的因果模型：一種用於有序離散數據，另一種用於無序離散數據。對於高維連續數據，我們提出了基於核的因果模型。最後，我們對隱共因子模型進行了改進。對於有序離散數據，我們將加性噪聲模型推廣到混合模型，允許存在多個函數。該模型可以處理數據從多個異構環境採集的情況，並且在每個環境中，果變量和因變量服從加性噪聲模型。為了推斷因果方向，我們依賴於一種新的統計不對稱性，這種不對稱性是根據聯合分佈的支持集來定義的。我們證明了p(effect|cause)的支持點數與p(cause|effect)的支持點相比變化較小，是因為p(cause|effect)的最小支持點數量非常小。這是由於函數的最小值和最大值通常只有很少的支持點。實驗結果也表明了該方法的有效性。另一種類型的離散數據是類別型數據，其沒有順序信息。通過從函數化模型轉變到非函數化模型，我們開發了一種有效的方法來處理這種情況，使用基於直接獨立性測試的新型統計不對稱。我們將p(cause)和p(effect|cause) 視為獨立的隨機變量，在因變量的每種可能的狀態下進行採樣，並使用距離相關係數來衡量它們的統計依賴性。距離相關係數在因果方向上較小。這是因為p(cause)與p(effect|cause)是獨立的，而p(effect)依賴於p(cause|effect)。我們進行了實驗和計算複雜性分析以顯示該推理方法的有效性和效率。在大數據時代，人們需要對高維變量進行因果推理。與上述情景不同，這裡定義統計不對稱應該避免估計聯合分佈。因此，我們使用核映射，並基於自由獨立條件給出新的統計不對稱性。自由獨立條件是在再生核希爾伯特空間中因變量的概率表示的協方差矩陣與給定因變量的果變量的條件概率表示的協方差矩陣自由獨立。這導致了因果不對稱，即設計的測量在因果方向上為零但在反因果方向上小於零，並且它揭示了因果方向。我們詳盡的研究了因果方向的可識別性，並且說明不可識別的條件難以滿足。可泛化的實證結果為我們的理論提供了證據。最後，我們從因果不對稱轉向不存在隱共因子與存在隱共因子的不對稱，以檢測因果之間的隱共因子。最近的研究假設模型係數為均勻採樣產生，冰使用回歸矢量產生的譜測度在不存在隱共因子與存在隱共因子的情況下的不對稱表現來識別線性模型中存在的隱共因子。譜測度在無隱共因子情況下是均勻的，但在隱共因子存在下是不均勻的。不同於比較譜測度的模式，我們計算一個名為一階矩度量的統計量。我們將其與均勻譜測度的一階矩度量進行比較。兩者在不存在隱共因子時是一致的，而在存在隱共因子的情況下彼此不同。這種統計的不對稱性可用於隱共因子檢測。我們的方法是省時的，並且可以在回歸矢量產生的譜測度不均勻實際情況下有效工作。我們在合成數據和實際數據上進行了實驗，以顯示所提方法的有效性和高效性。Liu, Furui.Thesis Ph.D. Chinese University of Hong Kong 2019.Includes bibliographical references (leaves 120-128).Abstracts also in Chinese.Title from PDF title page (viewed on 10, December, 2020).Liu, Furui
Steps Toward Flexible Speech Recognition
No full textThis paper describes recent progress at Tokyo Institute of Technology and the author's perspectives for making speech recognition systems more flexible at both the acoustic and linguistic processing levels. Specifically, it describes a broadcast news transcription system, a multimodal dialogue system for information retrieval, neural-network-based HMM adaptation for noisy speech, online incremental speaker adaptation combined with automatic speaker-change detection, message-driven speech recognition and understanding, speech summarization, a Japanese national project on spontaneous speech corpus and processing technology, and speech recognition in the ubiquitous/wearable computing environment. For processing spontaneous speech, indispensable will be a paradigm shift from speech recognition to understanding where underlying messages of the speaker are extracted, instead of transcribing all the spoken words. Building a large corpus of spontaneous speech to construct reliable acoustic and linguistic models is also crucial. Due principally to the technology of making computers smaller, more powerful and cheaper, the ubiquitous and wearable computing era is expected to come into being in the initial years of the 21st century. In such an environment, speech recognition will be widely used as one of the principal methods of humancomputer interaction
- …
