| 金川元信 | 1151039: M, 2回目発表 | 松本裕治, 池田和司, 新保仁, Kevin Duh |
|
title: Time series data analysis with kernel methods
abstract: Recently, Kernel methods to deal with probability distributions are developed and applied to wide range of problems such as two-sample problems, independence tests, conditional independence tests, etc. We apply kernel methods to estimate conditional distributions to filtering problems in state-space models. Experimental results show that our approach is better than existing methods to estimate hidden states. language of the presentation: Japanese | ||
| 小林龍 | 1151048: M, 2回目発表 | 松本裕治, 池田和司, 新保仁, Kevin Duh |
| 吉川友也 | 1151114: M, 2回目発表 | 松本裕治, 池田和司, 上田修功, 新保仁 |
|
title:Learning Information Diffusion Model with Latent Features of Users
abstract: Based on modeling diffusion and sharing of available piece of information on the social networks, predicting the phenomenon in the future and ining some novel knowledges useful are well-researched recently. In this presentation, we focus on the information diffusion phenomenon on the social networks, and propose a novel model to estimate the diffusion probabilities of he information diffusion model. In particular, we model that each node on the social networks has two types of latent features, then a diffusion probability of a piece of information for each ink is decided by the features which both ends of the link have. Therefore, we develop the parameter estimation method according to maximizing a posterior using EM algorithm. In experiments, we first show that the proposed model can estimate the model parameters more precisely than the existing models. Moreover, we can estimate how many nodes receive a certain piece of information using the proposed model accurately. language of the presentation: Japanese 発表題目: ユーザの潜在特徴を考慮した情報拡散モデルの学習 発表概要: ソーシャルネットワーク上の情報の拡散をモデル化することによって，将来の現象を予測したり，新しい知識の発見を行う研究が最近注目を浴びている． 本発表では，ソーシャルネットワーク上の情報拡散現象に焦点を当て，ユーザ間の情報の伝わりやすさを表す拡散確率を正確に推定するための確率モデルを提案する． 具体的には，各ユーザが二種類の潜在的な特徴を持つようにモデル化し，各リンクにおける情報の拡散確率はリンクの両端ノードのその特徴によって決まるようにする． 実験では，まず最初に，提案法が既存の方法よりも正確に拡散確率が推定できることを示す． また，推定した拡散確率を使って，何人に情報が伝わるかの指標である影響度をより正確に推定できることを示す． | ||
| 小嵜耕平 | 1161019: D, 中間発表 | 松本裕治, 池田和司, 新保仁, 小町守 |
|
title:
Graph construction with demographic queries for influence maximization
abstract: Influence maximization is the problem of finding a small fixed-length subset that maximize the spread of influence in social network. Kempe et al. shows the problem is NP-hard and the scalable solution is important for applying viral marketing in real-world dataset. In this study, we aim to develop a framework for maximizing the spread of influence to construct graphs by using demographic queries (name, sex, location). language of the presentation: Japanese 発表題目: セグメントをクエリとした影響最大化問題のためのグラフ構築 発表概要: 影響最大化問題は，情報が伝播する過程をモデル化した情報伝播モデルにおいて，影 響が最大となるようにサイズに制約を持つ初期の情報発信者集合を発見する，離散最適化問題である． この最適化は計算コストが大きいため，大規模な実データに対して効率的な計算をするためには高速化のための工夫が必要となる． そこで本研究ではセグメント（ユーザー層）をクエリとして，影響最大化問題のためのグラフ構築を行うことで，大規模な実データに対して現実的な計算手法を開発することを目標とする． | ||
| 木本奈津子 | 1261004: D, 中間発表 | 湊小太郎, 金谷重彦, 杉浦忠男, 田村俊世（大阪電通大） |
|
title: Improvement of measurement accuracy in pulse oximetry
abstract: Pulse oximeter measures the percutaneous arterial oxygen saturation (SpO2) by the absorption ratio of two different wavelength lights that pass through the tissue. However, sensor misalignment and incorrect position can cause inaccurate results. Especially in the case of neonates and infants, finger size or sensor location can create a gap between the sensor and finger. The gap may cause light leakage that does not pass through the tissue (“shunted light”). This might be one cause of inaccurate SpO2 measurement. We investigated the effect of shunted light on SpO2 value while displacing the sensor laterally from the center to the outside of the finger. We found that the sensor lateral displacement caused shunted light which degrades measurement accuracy. We also found that if the LED or PD is firmly attached to the finger and the shunted light is shielded, there were very few errors in the measurement value. We think that sensors which shield the shunted light and fit the finger firmly will improve measurement accuracy in pulse oximetry. language of the presentation: Japanese 発表題目: パルスオキシメータによるSpO2測定精度向上 発表概要: パルスオキシメータは，生体組織を透過する２つの異なる波長の減光度比によって経皮的動脈血酸素飽和度（SpO2）を計測する．しかしながら，センサの不適切な装着により測定精度が低下する．特に新生児や小児などの測定では，指のサイズや装着位置によってセンサと指の間に間隙が生じ，生体組織を通らない光が受光されてしまう（これを迷光とする）．我々は，センサを横方向にずらしながらSpO2測定を行い，迷光が測定値に与える影響について検討を行った．その結果，迷光によりSpO2値に誤差が生じることを確認し，また，迷光の遮光効果も確認した．加えて，LEDまたはPDのいずれか一方が指に密着固定されていれば，他方の素子が指中心からずれても測定値に生じる誤差は非常に少ないことがわかった．これらにより，指に密着する素材を使用し，遮光効果のある機構を加えたセンサの改良がSpO2測定の精度向上のために有用であると考えている． | ||
| 小林直樹 | 1261006: D, 中間発表 | 湊小太郎, 金谷重彦, 杉浦忠男, 田村俊世（大阪電通大） |
|
title: A study for
the mechanism of errors in pulse oximetry using a new electrical model applying
arterial blood inflow, tissue oxygen consumption and venous blood outflow. | ||
| NOR AINI BINTI ZAKARIA | 1261026: D, 中間発表 | 湊小太郎, 金谷重彦, 杉浦忠男, 田村俊世（大阪電通大） |
|
title: Application of Wearable Inertia Sensor in Clinical Practice abstract: This study reports the quantitative classification of falling risk among elderly using a single wearable inertia sensor, which combines accelerometer and gyrosensors devices, applied during standard physical assessment test: timed up and go test and also four square step tests. Subjects were categorized into two groups as low fall risk (LFR) and high fall risk (HFR) with total duration taken to complete the test as the threshold between them. One sensor was attached at the subject’s waist dorsally, and two sensors attached at right and left thigh. The acceleration and gyrosensor signals in three directions were extracted during the test. The analysis was carried out in phases to evaluate the test in each activity. Comparisons between the two groups using extracted parameters from signals could reveal the elderly activities in each phase. Classification using other parameters along with time parameter suggests that this is an improved method in evaluating fall risk, which promises benefits in terms of improvement of elderly quality of life. language of the presentation: English | ||
| 脇坂洋祐 | 1151118: M, 2回目発表 | 伊藤実, 安本慶一, 中島康彦, 柴田直樹 |
|
title:Task Scheduling Considering Dynamic Frequency Shift by TurboBoost and HyperThreading
abstract:We design an operating frequency model of Turbo Boost and Hyper-Threading Technology used in recent multi-core processors from actual measurements and propose a task scheduling algorithm considering a network contention and a dynamic frequency shift based on the designed model. Tasks are scheduled in order by using proposed algorithm. When decide each task's allocation to a processor, this algorithm generates the combinational permutation that satisfies the dependence among the tasks, schedule these permutation and select the better permutation of schedule result. In this time, the proposed algorithm consider the operating frequency and calculate the execution time of each task We compared the proposed algorithm and the existing task scheduling algorithms which consider the network contention and do not consider the dynamic shift of the operating frequency. The proposed algorithm reduced the total processing time by 19%. language of the presentation:Japanese 発表題目:ターボブースト・ハイパースレッディングによる動作周波数の動的変更を考慮したタスクスケジューリング 発表概要:近年のマルチコアプロセッサに実装されているターボブースト及びハイパースレッディングに着目し，実機実験を通した動作周波数モデルの構築と，このモデルを用いた動作周波数の動的変更及びネットワークコンテンションを考慮したタスクスケジューリングアルゴリズムを提案を行う．提案アルゴリズムでは，タスクをある順にスケジュールしていくが，各タスクをどのプロセッサに割り当てるか決定する際に，決められた数の依存関係を満たしたタスクの順列を生成し，これらの仮割り当てを行い，ベストのタスクを選択する．この際，動作周波数を考慮し，各タスクの処理時間の算出を行う．提案手法と，ネットワークコンテンションのみを考慮する既存のアルゴリズムをシミュレーションにより比較し，提案アルゴリズムが全体の処理時間を最大19%短縮できることを確認した． | ||
| 秋吉保紀 | 1151004: M, 2回目発表 | 井上美智子, 中島康彦, 畠山一実, 米田友和 |
|
title: IR-drop Estimation for LSI Test Patterns abstract: As shrinking feature sizes and speeding-up operating frequency of LSIs, the delay testing has become more important.The delay testing is usually performed with at-speed scan testing,in which the circuit is tested at functional frequency. However, at-speed scan testing is suffering from serious power issues. The switching activity in at-speed testing is much higher than that in functional mode. This may cause excessive IR-drop which significantly increases the signal propagation delay and in turn leads to IR-drop-induced timing failures. The amount of IR-drop depends on test patterns, that is, IR-drop analysis for each pattern is required to accurately validate whether a test pattern may cause IR-drop-induced timing failure.However, the IR-drop analysis is well known to consume large amount of time. Recently, a fast IR-drop estimation method has been proposed. The method effectively reduces the number of execution in IR-drop analysis by using power profile and IR-drop profile for a few representative patterns. This research proposes an extension of the IR-drop estimation method to further speed-up the entire estimation time while maintaining the estimation accuracy. The proposed method uses a fast toggle-count-based metric for power estimation instead of precise power analysis. As a result, the entire IR-drop estimation time can be effectively reduced. language of the presentation: Japanese 発表題目:LSIのテストパタンのIRドロップ見積り手法 発表概要:LSIの微細化,高速化によって遅延故障テストの重要性が増している.遅延故障テストでは一般に,回路を実動作速度で動作させるat-speedスキャンテストが実施される.しかしat-speedスキャンテストは通常動作時に比べてスイッチング数が多いため,深刻な消費電力問題に直面している.これが,重大な信号遅延を引き起こすような過度のIRドロップを発生させ,IRドロップ起因のタイミング違反を引き起こす可能性がある.また,IRドロップの量は,テストパタンに依存する.そのため,IRドロップ起因のタイミング違反の有無を検証するにはテストパタンごとにIRドロップ解析を行う必要がある.しかし,IRドロップ解析には膨大な実行時間が必要であることが知られている.最近,IRドロップを高速に見積る手法が提案された.この手法はいくつかの代表パタンの電力情報とIRドロップ情報を用いることでIRドロップ解析の実行回数を大幅に減らしている.本研究では既存手法の見積り精度を維持しつつ全体の実行時間をさらに短縮したIRドロップ見積り手法を提案する.提案手法は消費電力解析の代わりにより高速なトグルカウントを基にした電力見積りを行うことで,全体の見積り時間を大幅に短縮した. | ||
| 吉見優太 | 1151117: M, 2回目発表 | 井上美智子, 中島康彦, 畠山一実, 米田友和 |
|
title: Unknown values handling technique for in-field self test abstract: The semiconductor has the problem by deterioration. The deterioration cannot be detected by production test in a typical factory.Therefore, self testing in the field becomes important. However, self test has a problem that the compression of the test result is obstructed by the value that is called an unknown value, and it is important to evade this. The conventional technique handling unknown values are targeting production test, while it is necessary to keep constraints such as test data volume to apply them to in-field self test.This presentation shows the present progress of the study. language of the presentation: Japanese 発表題目:フィールドでのセルフテストにおける不定値処理手法 発表概要:半導体の劣化はシステムの信頼性に影響を及ぼす重大な問題である。しかし劣化は一般的な出荷時における生産テストにおいて検出することができない。そのためフィールドでセルフテストを行うことが重要となる。しかし、セルフテストにおいても不定値と呼ばれる値によってテスト結果の圧縮が阻害される問題があり、これを回避することが重要である。これに対する従来手法は対象が生産テストとなっており、フィールドテストで適用するためにはテストデータ量等の制約を守ることが必要である。本発表では、現在の進捗状況について報告する。 | ||
| Marcos Daniel Villagra | 1161015: D, 中間発表 | 中島康彦, 関浩之, 山下茂（立命館）, 中西正樹（山形大） |
|
Title: On the Power and Limitations of Quantum Computing Models
Abstract: To understand the physical limits of computation it is necessary to shift our classical computer models to ones that take into account physical considerations. The best current theory of physical reality is quantum mechanics, which take us to think on computer models based on it. Quantum computation is the study of the power and limitations of computer models that take into consideration quantum mechanical effects like interference and entanglement. Several of the classical computing models like boolean circuits, Turing machines, etc., can be extended to quantum models of computation. In this research we focus in two particular models: the decision tree or query complexity and communication complexity. This research is divided in two parts. First we consider upper bounds in the decision tree model. Here we study Quantum Walks, a very powerful paradigm for the design and analysis of quantum algorithms. Clear mathematical foundations are still lacking for this paradigm. Hence, as a step toward this objective, the following question is being addressed: Given a graph, what is the probability that a quantum walk arrives at a given vertex after some number of steps? This is a very natural question, and for classical random walks it can be answered by different combinatorial arguments. For quantum walks this is a highly non-trivial task. Furthermore, this was only achieved before for one specific coin operator (Hadamard operator) for walks on the line. Even considering only walks on lines, generalizing these computations to a general SU(2) coin operator is a complex task. The main contribution of this part is a closed-form formula for the question above for a general symmetric SU(2) operator for walks on lines. As a second contribution, we show how some basic properties of the walk can be deducted by means of weak convergence theorems for quantum walks. In the second part of this research we focus on lower bounds in communication complexity. In particular, quantum nondeterministic multiparty communication. There are three (possibly) different types of nondeterminism in quantum computation: i) strong, ii) weak with quantum proofs, and iii) weak with classical proofs. Here we focus on the first one. A strong quantum nondeterministic protocol accepts a correct input with positive probability, and rejects an incorrect input with probability 1. In this work we relate strong quantum nondeterministic multiparty communication complexity to the rank of the communication tensor in the Number-On-Forehead and Number-In-Hand models. In particular, by extending the definition proposed by de Wolf to nondeterministic tensor-rank (nrank), we show that for any boolean function f, 1) in the Number-On-Forehead model, the cost is upper-bounded by the logarithm of nrank(f); 2) in the Number-In-Hand model, the cost is lower-bounded by the logarithm of nrank(f). Furthermore, we show that when the number of players in the protocol is o(log log n) we have an exponential gap between strong quantum nondeterministic communication and 2-sided error quantum communication in the Number-On-Forehead model. language of the presentation: English | ||