Kyoto University (京都大学) / 0048 / 新制・課程博士 / 博士(工学) / 甲第15667号 / 工博第3325号 / 新制||工||1502(附属図書館) / 28204 / 京都大学大学院工学研究科材料工学専攻 / (主査)教授 乾 晴行, 教授 落合 庄治郎, 教授 白井 泰治 / 学位規則第4条第1項該当

京都大学 / 0048 / 新制・課程博士 / 博士(文学) / 甲第17764号 / 文博第623号 / 新制||文||593(附属図書館) / 30571 / 京都大学大学院文学研究科現代文化学専攻 / (主査)教授 伊藤 和行, 教授 林 晋, 准教授 伊勢田 哲治 / 学位規則第4条第1項該当 / Doctor of Letters / Kyoto University / DGAM

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第19294号 / 工博第4091号 / 新制||工||1631(附属図書館) / 32296 / 京都大学大学院工学研究科都市環境工学専攻 / (主査)教授 寶 馨, 教授 米田 稔, 教授 中北 英一 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

京都大学 / 0048 / 新制・課程博士 / 博士(農学) / 甲第20718号 / 農博第2247号 / 新制||農||1054(附属図書館) / 学位論文||H29||N5084(農学部図書室) / 京都大学大学院農学研究科森林科学専攻 / (主査)教授 神﨑 護, 教授 柴田 昌三, 教授 吉岡 崇仁 / 学位規則第4条第1項該当 / Doctor of Agricultural Science / Kyoto University / DGAM

在機器學習與深度學習技術帶動第三次人工智慧熱潮，特別是與機器人、大數據、3D列印等結合，「人工智慧」成為各大科技企業重點發展技術，無論是透過成立研究小組或是併購的方式，在2016年即有40多家人工智慧技術公司被併購，同時這些技術也被應用在各式產品與服務中。此外不同產業中也陸續引進人工智慧，從事需要耗時費力的基礎工作，節省成本，也引發人類被取代的恐慌。世界各國除著力投資發展人工智慧之外，也重視人工智慧為社會及經濟產生的影響與現行法制的衝擊。 人工智慧技術也應用於創作與發明過程中，且在機器學習技術下，人類僅需輸入指示與限制，人工智慧完成內容創作所產出作品，與人類創作成果並無二致。或是像是神燈精靈一般，人類只要以人工智慧可理解的方式定義問題、要件分析、功能設計，人工智慧即能完成最重要的物理設計，解決人類之問題，其產出物可能是符合產業利用性、新穎性與進步性等專利要件。這些人工智慧創作與發明物是否符合我國現行著作權法與專利法之規定而受到保護，將是本文探討重點，本文將以文獻研究以及比較法的方式深入研究。 智慧財產權制度設立的主旨是為保護人類精神活動成果，以人類為創作主體為前提，人工智慧參與創作之成果對於現行制度而言自然有所扞格。然而人工智慧創作力對於未來創作與發明而言，都是有所助益，可豐富文化的多樣性並加快技術的發展。而日本知識產權戰略本部也於2016年四月時也將針對人工智慧創作物之法律保護，擬修訂智慧財產權法。如我國未來亦研擬將人工智慧創作物納入法律保護，本文參考日本立法相關討論以及美國學者之見解，提出立法時應考量的權利歸屬以及衍生的相關問題。 / Machine Learning and Deep Learning are leading the new artificial intelligence era, especially when integrated with technologies of robot, big data, and 3D printing. As A.I. gradually became the one of the most popular technologies, corporate giants, such as Google, IBM, Facebook, and Apple, have been setting up research labs and acquiring A.I. startups to improve the quality of their services and products. Meanwhile, through using their service and products, our daily life is filled with A.I. Moreover, in many different industries, companies are using A.I. to reduce their cost by replacing labors from time-consuming jobs. Governments not only invest in the development of A.I. technology, but also response to the impact A.I. brings to the society, economic and Law. Artificial creativity is a new way for creation and invention. With machine learning, human only need to input the indication and limitation for A.I. to generate the outcome which is almost the same as what human can do. A.I. is also being described as a “genie in the machine”. Human input the description of their problems, functional analysis, and functional design, then A.I. will do the physical design to solve the problems and generate inventions which are useful, novelty and un-obviousness. Whether these creations and inventions are copyrightable and patentable is the core of this essay. Intellectual property system is aimed to protect the result from human’s mind activity, so the author or inventor must be human beings. When A.I. is not just a creation tool, but a creative subject, that’s where a conflict occurs. However, A.I. creativity is beneficial to human creative and inventing activity, because it can quicken the progress of technologies and improve culture diversity. Legislators in Japan are planning to protect A.I. creation through modification of intellectual property law. If we also expect to protect A.I. creation and invention in Taiwan in the future, with Japan legislative discussion and America scholars’ theories, this essay might offer some useful indications on the right attribution and other derivation problems.

本研究旨在以內容分析方式，分析台灣地區報紙有關墮胎新聞之報導，以瞭解閱聽人所得到的墮胎資訊為何。本研究選定中國時報與聯合報進行分析。分析結果發現台灣地區報紙對於墮胎新聞的報導，多數採用非個案報導，最常出現在墮胎新聞中的主角是未成年未婚者，多數的墮胎新聞是以恐懼手法呈現。對於防治墮胎所需之避孕與正確性教育與兩性關係資訊，台灣報紙在這些方面的報導極度不足。未來應加強此方面的資訊，以有效改善墮胎狀況。

風險偏好與預測精確性對生存力的重要性吸引進來許多理論學者的注意。一個極端是認為風險偏好完全不重要，唯一重要是預測精確性。然而此乃基於柏拉圖最適配置之下。透過代理人基模型，我們發現相異的結果，即風險偏好在生存力上扮演重要角色。 / The relevance of risk preference and forecasting accuracy to the survival of investors is an issue that has recently attracted a number of recent theoretical studies. At one extreme, it has been shown that risk preference can be entirely irrelevant, and that in the long run what distinguishes the agents who survive from those who vanish is just their forecasting accuracy. Being in line with the market selection hypothesis, this theoretical result is, however, established mainly on the basis of Pareto optimal allocation. By using agent-based computational modeling, this dissertation extends the existing studies to an economy where adaptive behaviors are autonomous and complex heterogeneous, and where the economy is notorious for its likely persistent deviation from Pareto optimality. Specifically, a computational multiasset artificial stock market corresponding to Blume and Easley (1992) and Sandroni (2000) is constructed and studied. Through simulation, we present results that contradict the market selection hypothesis. Risk preference plays a key role in survivability. And agents who have superior forecasting accuracy may be driven out just because of their risk preference. Nevertheless, when all the agents are with the same preference, the wealth share is positively correlated to forecasting accuracy, and the market selection hypothesis is sustained, at least in a weak sense.

運用語義網本體論來描述網路服務，實現網路服務的自動發現、調用和組合已經被證實是有效的。人工智慧規劃技術就是運用此技術描述網路服務來達到自動化網路服務組合。OWL-S支援使用OWL來描述網路服務的前提與效果，而在OWL的規則描述語言方面，則可以使用SWRL。本論文是以OntoComposer規劃工具為基礎，發展一套前置處理器來簡化其使用，讓使用者不需具備人工智慧規劃描述語言及語義網路服務描述等知識，只要尋找到符合需求之網路服務後，就可以自動轉換成支持條件分支圖規劃器之輸入文件，並在設定目標後進行規劃組合，最後讓組合之複合網路能夠在執行引擎上正確執行。 / Using Semantic Web ontologies to describe Web Services has proven to be effective for automatic service discovery, invokcation and composition. AI planning techniques have been employed to automate the composition of Web Services in this way. OWL-S supports the description of the preconditions and effects of a web service using OWL statements, and SWRL is the language for expressing OWL Rules. OntoComposer is an AI planning based tool for Combining-GraphPlan, an extension for GraphPlan so that supports condition branching. This thesis presents a preprocessor for OntoComposer to simplify its input task so that the user does not have to learn the AI Planning description language and knowledge of the semantic web service description. Just look to meet the demand of web services, our preprocessor will translate them to support branch planning for the input file. After setting the targets the OntoComposer will compose some component web services to a complex web service. Finally, let the composition of web services on the execution engine the correct execution.

各類電子商務網站上的推薦機制應用已日趨廣泛且成熟。而隨著決策問題日漸複雜，現行的推薦機制發展已經可以看到應用的界限，再也無法貼近使用者所面臨的複雜問題。現行的推薦機制架構需要被重新審視、定義與設計其核心演算法。本研究用更寬廣的角度看待推薦機制，並將改良後的推薦機制視為解決問題的新典範。 首先，本研究定義了改良後的推薦機制所應支援的功能，包括：階層式條件的多維度推薦以及多階段的推薦。多維度推薦機制能夠讓使用者從不同的面向去看待決策問題，而階層式條件則允許使用者針對每個維度再往下設定階層式條件，幫助決策者更貼切的描述所遭遇的問題，如此一來推薦機制所提供的推論結果才能更符合決策者的原意。而多階段推薦則是協助決策者進行一連串的規劃方案，而這樣的推薦結論能夠提供可行方案的遠景，讓決策者能夠預先為可能發生的狀況進行準備，進而深化決策者對目前推薦結論的信心。 除了力求每個(或多個)階段推薦結論的正確性，推薦系統也要與所有的決策階段緊密結合（不僅止於資料搜集階段），所以必須能夠提供決策者行為面的建議，確切的建議決策者應該採取的行動。確切的行為面資訊推薦結論對於決策活動的參考價值更高。 所以，本研究修改了傳統的案例推導法(CBR)，試圖讓傳統CBR演算法成為符合改良後個案推薦機制的規範，因為CBR演算法最符合人類求解問題的邏輯程序，因此本研究在改良式個案推薦機制中重現CBR演算法中的4R推理循環。而且為了真正落實修正後的CBR演算法，本研究還結合了基因演算法提出GCBR的概念，幫助改良式個案推薦系統能夠更快速有效的收斂出推薦的結論。 最後，本研究也預期所提出的推薦機制能夠應用於各種不同的領域，而為了驗證所提出的推薦機制執行效率與可行性，本研究也列舉了數個實驗進行的規範方案。本研究所提出的改良式個案推薦機制核心演算法為一概化模型，能夠求解不同型態的決策問題。 / Recommender system can be regarded as fundamental technology of electronic commence web site. Some researchers also claimed that recommender system push the electronic web site to another development peak. Recommender system would need some mechanisms. These recommender mechanisms should be reviewed, redefined and expanded to include particularly case-based mechanism that focus on reality problem solving. Recently, CBR applications had been extended to provide recommendation mechanism based on previous cases. The abstract recommendation problems are usually hard to be formulated in strict mathematic models, and often solved via word-mouse experience. Case-Based Reasoning (CBR) is a paradigm, concept and instinctive mechanism for ill-defined and unstructured problem solving. Similarly to human problem solving process, CBR retrieves past experiences to reuse for target problem. Of course, the solutions of past cases may need to be revised for applying. The successful problem-solving experiences are then retained for further reusing. These are well-known 4R processes (retrieve, reuse, revise, and retain) of traditional CBR. Nevertheless, the case-based recommender mechanism is particularly suitable for reality problem reference because case-style can be used to describe unstructured problem. The next generation recommender mechanism should focus on the real life problem solving and applications. Thus, case-based recommender mechanism can be regarded as a new problem solving paradigm. To enhance traditional CBR algorithm to case-based recommender mechanism, the original CBR should be redesigned. In the traditional CBR algorithm, based on multiple objectives, the retrieved cases could provide to decision maker for references. However, as the decision problem is getting complex, pure multiple objective problem representation is too unsophisticated to reflect reality. Thus, a revised CBR algorithm equipped with capability to deal with more complexity is needed. Additionally, decision makers would wish to achieve the actionable information. The existing recommender mechanism can not provide the actionable direction to decision maker. Based on previous cases provided by CBR, decision maker would further hope that recommender mechanism could tell them how to do. These capabilities should be included into traditional CBR algorithm. Furthermore, traditional CBR has to evaluate all cases in case base to return the most similar case(s). The efficiency of CBR is obviously negatively related to the size of case base. Thus, a number of approaches have devoted to decrease the effort for case evaluation. This research proposes a revised CBR mechanism, named GCBR, which can be regarded as next generation CBR algorithm. GCBR can be applied to reality applications, particularly case-based recommender mechanism. Thus, it can be treated as a new problem solving paradigm. It also intends to improve traditional CBR efficiency stability no matter what kinds of case representation and indexing approaches.

就目前所知，這是第一篇應用人工類神經網路在股票風格投資方面的研究。類神經網路在樣本內與樣本外的分類正確率皆優於區別分析，而且類神經網路在樣本內的訓練範例中達成了百分之百的分類正確率。此外，我們也解決了傳統方法無法展示股票風格動態的問題。 檢視各種風格投資策略在台灣股票市場的績效表現之後，我們以神經網路為基礎，提出一個簡單而容易實行的投資策略。由這個策略的表現可以說明，即使在考慮了風險因素之後，積極的風格投資策略的確可以增加投資組合的績效表現。 / This is the first study of applying artificial neural networks (ANN) to classify and identify the equity styles. Regarding the accuracy, ANN outperforms discriminant analysis (DA) in all pure samples from 1987 to 1997. The ANN also commits the 100% classification accuracy for the in-sample training samples. In addition, the problem that traditional approach couldn't show equity style dynamics was solved with ANN and DA. The performances of style investing strategies were examined in Taiwan stock market. The proposed strategy is easily implemented by constructing portfolios based on the return, which neural networks forecasted. There is good evidence to show this simple strategy could enhance profit on the return and risk adjusted basis. This gives one evidence to illustrate that active style investing would add value.