16775 research outputs found
Sort by
Electrochemical Reduction of CO2 by SnS⎸PTFE⎸Pt Surface in an Aqueous Imidazole Medium: Catalysis and Kinetics
Get PDFDevelopment of a novel SnS⎸PTFE⎸Pt electrode as an efficient catalyst for electrochemical CO2 reduction to formate in aqueous alkaline imidazole medium is reported in this paper. The electrode was prepared through the incorporation of poly-tetrafluoroethylene (PTFE) with SnS and modification of Pt surface by SnS/PTFE. The electrode facilitated a first-order CO2 reduction at -1.1 V vs. Ag/AgCl (saturated KCl) via a diffusion-controlled pathway with a cathodic electron transfer coefficient ( ) of 0.22 and two-electron transfer kinetics in two steps, where the first electron transfer step determined the reaction rate. The SnS⎸PTFE⎸Pt electrode surface exhibited well-defined cyclic voltammograms (CVs) for CO2 reduction in comparison to GCE, Au, Pd, and graphite-modified electrodes utilizing the same catalyst, indicating superior catalytic efficiency and its conduciveness to kinetic investigations. Such performance indicates that the SnS⎸PTFE⎸Pt electrode would be a promising alternative for CO2 reduction in the context of renewable energy generation
視野による制約を前提としたゲームAIの強化学習
No full textゲームAIは主に強さの向上を目的として研究されてきたが，近年人間らしい挙動の獲得も研究が進められている．本研究では，人間の視野特性に着目し，強化学習エージェントに動的な視野を実装することで，「視野の制約を前提とした行動」の再現を目指す．提案手法では，注視点の動的な決定と注視点からの距離に応じた情報のぼかしを導入し，エージェントがゲーム内行動と視野移動を同時に学習する．Atari Breakoutに簡単な視野の制約を導入した実験では，視野の移動を含めた学習が可能であることを確認した．さらに視野の制約がある場合に，序盤の成績がかえって向上する傾向を確認した．今後は，格闘ゲームなどのより複雑なゲーム環境での学習の確認や，「視野の制約を前提とした行動」の再現を目指す．While game AI research has primarily focused on improving playing skills, recent studies have also explored the acquisition of human-like behavior. This research aims to reproduce “actions based on visual field constraints” by implementing a dynamic visual field in reinforcement learning agents, focusing on human visual characteristics. The proposed method introduces dynamic determination of the gaze point and information blurring based on distance from the gaze point. In the system, the agents simultaneously learn in-game actions and visual field movement. Experiments with simple visual field constraints in Atari Breakout confirmed the feasibility of learning even with visual field movement. Furthermore, we observed a tendency for improved scores in the early stages when visual field constraints were present. Future challenges include verifying learning in more complex game environments, such as fighting games, and reproducing actions based on visual field constraints.第29回ゲームプログラミングワークショップ (GPW-24), 箱根セミナーハウス, 2024年11月15日-17
組織内自尊感情の醸成を目的とした敬意伝播支援システムに関する基礎的検討
No full text本研究では，組織において各成員が好ましい組織内自尊感情を持てるようにすることが組織の活動の促進・向上に有益であり，そのためには成員同士が相互に敬意を表明しあえる環境を構築することが必要かつ重要であるという観点に立ち，そのような環境の実現を目指している．この最終目標に向けて，組織成員間で敬意メッセージを伝えあいやすくすることを狙った2 つのプロトタイプシステムを実装した．本稿では，この2 つのプロトタイプシステムの詳細と，それらを用いた実験を通じてその有効性を検証した結果について述べる．結果として，組織内で各成員が他成員に対する敬意情報を発信するオンラインメディアを用意し，そこで発信された情報をたまり場で提供する機能を有する，人の注意を惹きやすい特徴を有する物理的オブジェクトを設置する手段をとることにより，組織内での敬意情報のやり取りが活性化される可能性が示された．It is beneficial to enable each member within an organization to have a favorable feeling of self-esteem for promoting and improving organizational activities. Therefore, we aim to create an environment where the organization's members can readily express mutual respect. Toward this end, we have implemented two prototype systems that facilitate to exchange of mutual respect messages among organizational members. In this paper, we describe the details of the two prototype systems and verify their effectiveness through experiments using them. As a result, the possibility of activating the exchange of respect information within an organization was demonstrated by preparing online media to submit respect information, and by installing a physical object that has features to attract people's attention easily and that has the function of providing the information submitted in the online media at a hangout. In total, it was shown that the proposed method has the possibility of activating the exchange of respectful information within an organization
Softlabel Classification for Multimodal Sentiment Estimation using Multiple Third-party Annotations
Get PDFSupervisor: 岡田 将吾先端科学技術研究科修士(情報科学
Identifying Underlying Logic of Fallacious Arguments via Structured Templates
No full textSupervisor: 井之上 直也先端科学技術研究科修士(情報科学
