16775 research outputs found
Sort by
分子線エピタキシー(MBE)法により成長したGaAs(111)B上MnAs/InAs/MnAsダブルヘテロ構造の縦型スピンデバイス応用
Get PDFSupervisor: 赤堀 誠志先端科学技術研究科博
Anti-modularization for both high robustness and efficiency including the optimal case
No full textAlthough robustness of connectivity and modular structures in networks have been attracted much attentions in complex networks, most researches have focused on those two features in Erdos-Renyi random graphs and Scale-Free networks whose degree distributions follow Poisson and power-law, respectively. This paper investigates the effect of modularity on robustness in a modular d-regular graphs. Our results reveal that high modularity reduces the robustness even from the optimal robustness of a random d-regular graph in the pure effect of degree distributions. Moreover, we find that a low modular d-regular graph exhibits small-world property that average path length is O(logN). These results indicate that low modularity on modular structures leads to coexistence of both high robustness and efficiency of paths
ゲームの展開に過度に影響されないプレイヤ強さの推定方法
No full text近年ではゲームAIを用いて人を楽しませたり指導する研究が進められている．そういった中で，プレイヤの棋力を正確に推定することは，適切な対戦相手AIの用意やプレイヤの実力向上のためのフィードバックなどに活用できる．チェスや将棋，囲碁においては強いゲームAIの最善手とプレイヤの着手の評価値の差分「損失」を使った棋力推定が行われている．しかし，損失はゲームの展開に大きく影響されてしまい，少ない棋譜では推定結果がばらついてしまうという課題がある．本研究ではそういったゲームの展開による過度な影響を抑制した指標の作成方法を提案する．具体的には，連続して損失が高い手が出た場合にその一部をカウントしない，ばらつきの大きい試合後半を見ない，展開に影響されにくい「形の良さ」を見る，といった工夫を行った．これをもとに少数棋譜に対して棋力推定を行い，提案手法によって棋力の推定結果の標準偏差を0.387 としつつ精度をRMSE 0.752とすることができた．In recent years, researchers has used game AI to entertain and teach human players. In this context, accurate estimation of players’ strength is crucial for preparing appropriate opponent AIs and providing feedback for skill improvement. In chess, shogi, and Go, researchers have used the metric “loss,” which is the difference in evaluation values between strong game AI’s best move and the player’s move, to estimate players’ strength. However, there is a problem that such losses can be very much influenced by game progresses, and the results of players’ strength estimation vary with a small number of game records. In this paper, we proposed methods for calculating strength evaluation metrics that reduce excessive influence from game progresses. Specifically, we applied the following methods: not counting some high-loss moves when they occur consecutively, not using the second half of the game, which has a large variation , and using the metric ”goodness of shape”, which is less influenced by game progresses. We applied filters to the calculation, such as not counting some high-loss moves when they occur consecutively. Based on this, we estimated players’ strength using a small number of game records. As a result, we succeeded in achieving a standard deviation of 0.387 and RMSE 0.752 using our porposed methods.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
Effects of Alkyl Side Chain Length on the Structural Organization and Proton Conductivity of Sulfonated Polyimide Thin Films
No full textThis study investigates the impact of alkyl side chain length on structural organization and proton conductivity of sulfonated polyimide (SPI) thin films. SPIs with different alkyl sulfonated side chain lengths CX (X: number of carbon atoms at the side chain, X = 0, 3, 6 and 10) were synthesized and the relationship between molecular architecture and proton transport properties was investigated. In all SPIs, the polymer backbone oriented parallel to the substrate, and the lamellar structures were confirmed, with spacing increasing linearly with side chain length. Water uptake behavior and proton conductivity varied significantly, where C0 thin film exhibited the highest water uptake and proton conductivity at lower humidity. While, C3 thin film achieved higher conductivity under high humidity, reaching 1.8 × 10−1 S cm−1 at 298 K and 95% relative humidity (RH) and the activation energy of 0.18 eV at 90% RH. Conversely, extending the alkyl side chain length led to the insolubility in water in C10. As a result, a proton exchange membrane system with high chemical/water stability and high proton conductivity was constructed. These findings provide critical insights into designing advanced proton-conducting materials, optimizing their performance for fuel cells and other electrochemical devices
