In English vocabulary learning, continuation is an important factor; however, many learners are not good at continuing learning because they tend to prefer amusement or rest. Our proposed system targeting who eager learn but able continue for various reasons. We especially focused on and described approach have difficulty with learning. developed application aggressively supports the learners' sustainable motivation by gamification techniques efficient setting method.

The development of deep learning has led to the proposal various models for human activity recognition (HAR). Convolutional neural networks (CNNs), initially proposed computer vision tasks, are examples applied sensor data. Recently, high-performing based on Transformers and multi-layer perceptrons (MLPs) have also been proposed. When applying these methods data, we often initialize hyperparameters with values optimized image processing tasks as a starting point. We suggest that comparable...

本研究はパズルゲーム『ぷよぷよ』において深層強化学習を適用したものである.従来のルールベースの手法や関連性行列を用いた手法では,人間のトッププレイヤーが構築するような大きい連鎖を構築することができないという課題があった.また,深層強化学習による先行研究も,複雑な戦略を学習することが難しく,十分な性能を示せていないのが現状である.本研究では,深層強化学習によるぷよぷよAIの性能向上を目的とし,並列Actorと優先度付き経験再生を用いた.提案手法を評価するために,自作のぷよぷよ環境を用いて実験を行った結果,提案手法は平均最大連鎖数6.243,平均スコア33114を達成し,従来の深層強化学習による研究を上回る性能を示した.

The feature extraction of human activity recognition (HAR) based on sensor data has been studied as a hand-crafted method. significant ability is key factor in improving the accuracy HAR. Recently, deep learning methods have employed for extraction. In this paper, we review previous studies HAR and discuss suitable models First, applied various convolutional neural networks to clarify effective architecture Afterward, developed advanced by embedding submodules, such self-attention recurrent...

Deep learning (DL) models for sensor-based human activity recognition (HAR) are still in their nascent stages compared with image recognition. HAR's inference is generally implemented on edge devices such as smartphones because of its secure privacy. However, lightweight DL HAR, while meeting the hardware limitations, lacking. In this study, using neural architecture search (NAS), we investigated an effective model that can be used smartphones. We designed multiple spaces type convolution,...

Although many researchers have widely investigated activity recognition using smartphone sensing, estimation accuracy can be adversely affected by individual dependence. The result of our survey showed that the process sensor based has not been sufficiently discussed, especially representation learning Convolutional Neural Network (CNN). effectiveness model CNN in was verified, as were 10 types models: Deep (DNN) Hand-Crafted (HC) features, simple model, AlexNet, SE-AlexNet, Fully (FCN),...

In this study, we proposed the method of inference genetic networks which expresses regulation genes. The does not solve differential equations, learns using product-unit-neural-network (PUNN) and infer S-system model describes a set equations. experimental results show proposal is 160 times faster than previous estimated while maintaining equivalent performance to method.

In this study, we propose a system to determine the placement of smartphone using acoustic properties surface materials nearby. Detecting surrounding allows change notification method automatically based on situational factors. Researchers have studied how recognize position in which smartphones are worn while walking, accelerometer data; however it is difficult identify smartphone's stationary, because value does not significantly when put down. developed close echoes, an assumption that...

Abstract Particle swarm optimization (PSO) is a population‐based stochastic technique, inspired by the social behavior of birds (flocking) or fish (schooling), which applied to various problems in nonlinear systems. The inertia weights approach (IWA) and constriction factor (CFA) are improved methods PSO. IWA searches problem space globally early steps, finally locally near optimal solution. CFA method that introduces new parameter into velocity update equation. This paper proposes...

To provide cultural night experiences with tourists in Kanazawa City, we exhibited our interactive digital installation titled "Temari and Shadow" at the open space of 21st Century Museum Contemporary Art, Kanazawa. Our utilizes shadows cast by people standing front a projector. We chose Kanazawa's Kaga Temari, hand-sized colorful toy ball made thread, as motif this work. shadow-based system achieved 60 fps rendering estimate that time required for response shadow is 133 ms. However, latency...

In this study, we improved the usability of smartphones by automating a user’s operations. We developed an intelligent system using machine learning techniques that periodically detects context on smartphone. selected Android operating because it has largest market share and highest flexibility its development environment. paper, describe application automatically adjusts volume. Adjusting volume can be easily forgotten users need to push buttons alter depending given situation. Therefore,...

This paper proposes a system to determine the placement of smartphone by using acoustic properties surface materials nearby. Detecting surrounding allows change its notification method automatically, based on situational factors. Researchers have studied how recognize position in which smartphones are worn while walking, accelerometer data; however, it is difficult identify smartphone's stationary, because value does not significantly when put down. In this paper, we developed close...

産業用画像における異常は論理的な異常と構造的な異常に分類される.論理的な異常は物体の欠損や個数の過剰,配置のミスの総称であり,一方で構造的な異常は汚れや傷,異物の混入などを指す.従来の正規化フローによる異常検知手法では,特徴マップの局所的な情報を加味して変数変換が行われる. これらの手法は一般に,構造的な異常に関しては検出性能が高い一方で,論理的な異常の検出を苦手とする.本研究では,この問題に対処するために,正規化フローに自己注意機構を導入して変数変換時に特徴間の関係性を捉えるMulti-head Self Attention Flow(MSAFlow)を提案する.異常検知のベンチマークデータセットであるMVTecLOCOを用いて,提案手法と従来の正規化フローを比較した結果,全カテゴリの平均AUROCで5%の性能向上を達成した.

Detecting anomalies such as incorrect combinations of objects or deviations in their positions is a challenging problem industrial anomaly detection. Traditional methods mainly focus on local features normal images, scratches and dirt, making detecting the relationships between difficult. Masked image modeling(MIM) self-supervised learning technique that predicts feature representation masked regions an image. To reconstruct regions, it necessary to understand how composed, allowing within...

Although the volume settings of smartphones are important for users, they still need to push hardware sound button manually. The purpose our study is improve usability settings. Our proposed method predicts user's routine by learning actual daily smartphone logs. Related works used suitable input experimental participants learn setting pattern each user. In contrast, this uses This paper describes three results analyses many First, we investigate rate at which users change application...

By creating content that visualizes the historical landscape resources of Kanazawa City, it has become possible to visually grasp changes in landscape. Besides, by visualizing hidden and communicating those an easy-to-understand manner, tourists residents can learn more about history Kanazawa, will be opportunity enhance appeal excursion city. We developed a sightseeing application which was approved successfully released on March 20, 2020, for App Store so we could gather verify opinions at...

頻発激甚化する自然災害による被害を低減するためには住民自らが常日頃から災害渋滞時の避難行動を学習し,適切な避難経路を考える必要がある.本研究では,災害時に生じる渋滞を考慮した避難経路を学習できる「防災すごろく」のアプリケーション(アプリ)を開発した.本アプリではランダムな出発点から避難所に向かう間に豪雨,火災,渋滞の発生に伴う避難方法や避難経路の変更を学ぶことができる.また,ゲーム性を考慮し,グループ同士で得点を競うことや防災クイズを実施することで学習意欲の向上を図った.更に,災害疑似体験ならびにゲームに対する慣れを軽減するためにVRコンテンツを用いた災害表現を行っている.本アプリによる防災教育を行い,アンケートを実施した.その結果,本アプリの学習効果が示されると共に課題も明らかになった.

This paper presents Multi-State Real-Time Bidirectional Search (MSRTBS), a method that improves the efficiency of heuristic search algorithm for finding approximate solutions. A* (RTA*) is representative The Commitment (MSC) was introduced into RTA* and dramatically improved performance in problems such as N-puzzle. As well, (RTBS) also by changing unidirectional bidirectional one. proposed introduces MSC RTBS. experimental results showed compared with RTBS our method, MSRTBS, executed time...

In this paper, we applied various optimal solution search methods from the field of artificial intelligence to multiple sequence alignment, which is a basic problem analysis in bioinformatics, conduct performance evaluation experiments. The techniques were A*, Enhanced Iterative Deepening A* (IDA *), Recursive Best First Search (RBFS), MREC, and Partial Expansion (PEA *). Although MREC had been considered ineffective for 15-puzzle since it could only reduce number states by mere 1% compared...

A hypothetical reasoning, which can find an explanation for a given set of observation by assuming some sets, is useful knowledge processing framework because its theoretical basis and usefulness practical problems such as diagnostics, scheduling, design. It is, however, known to be computationally very expensive large it must deal with incomplete knowledge. Predicate-logic allows powerful compact representation compared propositional-logic. Efficient methods proceed reasoning have been...

In this study, we propose State Magic, which estimates the smartphone state using various sensors as standard equipment. Smartphone means location such placing it in a pocket or putting on desk. If can estimate its own state, developers create consumer support applications, an application for preventing mis-operations while is stored pocket. Six test participants operated some action tasks to measure sensor values our experiments. The experimental results show that method classify six...

コミュニケーションにおいて，言葉以外の感情などの表現は重要な役割を果たしており，計算機による感情認識の研究が行われている．本研究では，短時間の強い感情であり行動や生理の変化などに表出される情動について認識を試みた．一般的な服と変わらず扱うことができる感圧導電性衣服を着用することで，非言語的行動である身体の動きを計測した．感情を快適性と覚醒性の２つの軸で表すラッセルの円環モデルにより感情を記録した．計測した身体の動きと記録した感情から，機械学習による感情の認識を行った．