Search Results 1-20 of 518

  • Scheduling Computation Graphs of Deep Learning Frameworks for Multi-core CPUs  [in Japanese]

    樋口 兼一 , 田浦 健次朗

    Performance hotspots of such frameworks are calculations in each node and the existing work has focused on speed-up them. … Then we expect there is a speeding up of inference and learning in such networks. … The system is built on Chainer, and evaluations for overall performance improvement including speeding up on each node are conducted in several models. …

    情報処理学会論文誌プログラミング(PRO) 13(1), 20-20, 2020-01-29

    IPSJ 

  • Goal-Directed Behavior under Variational Predictive Coding: Dynamic organization of Visual Attention and Working Memory

    Minju Jung , Takazumi Matsumoto , Jun Tani

    … For this purpose, we propose a neural network model based on variational Bayes predictive coding, where goal-directed action planning is formulated by Bayesian inference of latent intentional space. … Furthermore, our analysis of comparative experiments indicated that the introduction of visual working memory and the inference mechanism using variational Bayes predictive coding significantly improved the performance in planning adequate goal-directed actions. …

    2019 IEEE/RSJ International Conference on Intelligent Robots and Systems, 1040-1047, 2020-01-27

    IR 

  • Determinants of Tax Revenue Performance in Malawi : Evidence of Direction and Dynamic Inference by ARDL Modelling

    Chilima Isaac Y.

    Numerous studies have identified significant relationships between macroeconomic variables and taxes. Noting however that the direction of 'causality' is not so common between countries, this study so …

    横浜国際社会科学研究 24(3), 53-75, 2020-01-20

    IR  DOI 

  • Determinants of Tax Revenue Performance in Malawi : Evidence of Direction and Dynamic Inference by ARDL Modelling

    Chilima Isaac Y.

    横浜国際社会科学研究 = Yokohama journal of social sciences 24(3), 317-339, 2020-01

  • Countermeasure against Backdoor Attack on Neural Networks Utilizing Knowledge Distillation

    Yoshida Kota , Fujino Takeshi

    … A DNN model that is trained with the tampered training dataset can achieve a high classification accuracy for clean (normal) input data, but the inference on the poisonous input data is misclassified to the adversarial target label. … Experimental results showed that the distilled model achieves high performance equivalent to that of a clean model without a backdoor.</p> …

    Journal of Signal Processing 24(4), 141-144, 2020

    J-STAGE 

  • Pilot Decontamination in Massive MIMO Uplink via Approximate Message-Passing

    FUJITSUKA Takumi , TAKEUCHI Keigo

    … The other is joint channel and data estimation via approximate message-passing (AMP) for bilinear inference. … The convergence property of conventional AMP is bad in bilinear inference problems, so that adaptive damping was required to help conventional AMP converge. … Numerical simulations show that the proposed AMP outperforms conventional AMP in terms of estimation performance when adaptive damping is not used. …

    IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences, 2020

    J-STAGE 

  • Improved Meta-learning by Parameter Adjustment via Latent Variables and Probabilistic Inference  [in Japanese]

    SHIMIZU Eiki , AOKI Shogo , MIKAWA Kenta , GOTO Masayuki

    … <p>Standard deep neural networks require large training data and fail to achieve good performance in the small data regime. … In this paper, we propose a model that adjusts learning rate for each task by introducing latent variables and applying probabilistic inference. … We demonstrate that this approach improves the performance of MAML on few-shot image classification benchmark dataset, and confirm that learning rate is adaptively adjusted by visualizing latent variables.</p> …

    Proceedings of the Annual Conference of JSAI JSAI2020(0), 4I3GS202-4I3GS202, 2020

    J-STAGE 

  • A role of causal inference in non-stationary environments  [in Japanese]

    SHINOHARA Shuji , OKAMOTO Hiroshi , MANOME Nobuhito , SUZUKI Kouta , MITSUYOSHI Shunji , CHUNG Ung-il

    … <p>Bayesian inference is a process of narrowing down hypotheses (causes) to one that best explains observational data (effects). … However, the object of inference is not always constant. … We propose an extended Bayesian inference (EBI), incorporating human-like causal inference into Bayesian inference and evaluate the estimation performance of EBI through the simple numerical experiments on coin toss. …

    Proceedings of the Annual Conference of JSAI JSAI2020(0), 3Rin413-3Rin413, 2020

    J-STAGE 

  • Amortized Variational Inference for Sets  [in Japanese]

    TANIGUCHI Shohei , IWASAWA Yusuke , MATSUO Yutaka

    … <p>Amortized variational inference is a powerful tool to approximate intractable posterior distributions over local latent variables using parametric functions (e.g., neural networks). … In this paper, we study amortized variational inference for permutation-invariant sets rather than fixed dimensional vectors, which has a wide range of applications, such as uncertainty-aware regression and image completion. …

    Proceedings of the Annual Conference of JSAI JSAI2020(0), 2D4OS18a03-2D4OS18a03, 2020

    J-STAGE 

  • Computational Semantics for Comparatives based on CCG and Theorem Proving  [in Japanese]

    HARUTA Izumi , MINESHIMA Koji , BEKKI Daisuke

    … <p>Comparative constructions pose a challenge to Natural Language Inference (NLI), a task of determining whether a text entails a hypothesis. … However, a computational inference system for comparatives is not developed enough to be used for NLI tasks. …

    Proceedings of the Annual Conference of JSAI JSAI2020(0), 1E3GS905-1E3GS905, 2020

    J-STAGE 

  • Does increasing an athletes' strength improve sports performance? A critical review with suggestions to help answer this, and other, causal questions in sport science

    Steele James , Fisher James , Crawford Derek

    … One area of interest is the role that muscular strength, and thus approaches to improve this (i.e. resistance training), has upon sports performance. … In this review we briefly consider the evidence regarding an answer to the causal question "<i>Does increasing an athletes' strength improve sports performance?</i>". …

    Journal of Trainology 9(1), 20, 2020

    J-STAGE 

  • Recurrent Neural Network Compression Based on Low-Rank Tensor Representation

    TJANDRA Andros , SAKTI Sakriani , NAKAMURA Satoshi

    … But most of these RNNs require much computational power and a huge number of parameters for both training and inference stage. … First, we evaluate all tensor-based RNNs performance on sequence modeling tasks with a various number of parameters. … Based on our experiment result, our proposed TT-format GRU are able to preserve the performance while reducing the number of GRU parameters significantly compared to the uncompressed GRU.</p> …

    IEICE Transactions on Information and Systems E103.D(2), 435-449, 2020

    J-STAGE 

  • Bayesian Inference for Mixture of Experts Using Replica Exchange Monte Carlo Method  [in Japanese]

    松平 京介 , 永田 賢二 , 本武 陽一 , 岡田 真人

    … The purpose of this paper is to realize the Bayesian inference for Mixture of Experts (ME) with replica exchange Monte Carlo (REMC) method, and to evaluate the singular structure of ME by analyzing posterior distribution through numerical simulation. … In the singular statistical model, The Bayesian estimation is superior to the Maximum likelihood method from the viewpoint of generalization performance. …

    情報処理学会論文誌数理モデル化と応用(TOM) 12(3), 37-45, 2019-12-23

    IPSJ 

  • Attitude Detection for One-Round Conversation: Jointly Extracting Target-Polarity Pairs

    Zhaohao Zeng , Ruihua Song , Pingping Lin , Tetsuya Sakai

    … Our experimental results show that treating the two subtasks independently is not the optimal solution for Attitude Detection, as achieving high performance in each subtask is not sufficient for obtaining correct target-polarity pairs. … By employing pointer networks to consider the target extraction task a boundary prediction problem instead of a sequence labelling problem, the model obtained better performance and faster training/inference than LSTM and LSTM-CRF based models. …

    情報処理学会論文誌データベース(TOD) 12(4), 2019-10-23

    IPSJ 

  • A Novel Predictive-Coding-Inspired Variational RNN Model for Online Prediction and Recognition

    Ahmadreza Ahmadi , Jun Tani

    … Its architecture attempts to address two major concerns of variational Bayes RNNs: how latent variables can learn meaningful representations and how the inference model can transfer future observations to the latent variables. … We demonstrate better prediction performance on a robot imitation task with our model using error regression compared to a standard variational Bayes model lacking such a procedure. …

    Neural Computation 31(11), 2025-2074, 2019-10-17

    IR 

  • Reconstructing neuronal circuitry from parallel spike trains

    Kobayashi Ryota , Kurita Shuhei , Kurth Anno , Kitano Katsunori , Mizuseki Kenji , Diesmann Markus , Richmond Barry J. , Shinomoto Shigeru

    … The performance of inference is optimized by counting the estimation errors using synthetic data. …

    Nature Communications (10), 2019-10-02

    IR 

  • Equality tests of high-dimensional covariance matrices under the strongly spiked eigenvalue model

    矢田 和善 , 青嶋 誠 , Aki Ishii , Kazuyoshi YATA , Makoto AOSHIMA

    … We discuss the performance of the test procedure by simulations. …

    Journal of statistical planning and inference (202), 99-111, 2019-09

    IR 

  • On-Device Deep Learning Inference for Efficient Activity Data Collection

    Mairittha Nattaya , Mairittha Tittaya , Inoue Sozo

    … The performance of the systems greatly depends on the quantity and "quality" of annotations; … While mobile and embedded devices are increasingly using deep learning models to infer user context, we propose to exploit on-device deep learning inference using a long short-term memory (LSTM)-based method to alleviate the labeling effort and ground truth data collection in activity recognition systems using smartphone sensors. …

    Sensors 19(15), 3434-1-3434-20, 2019-08-05

    IR 

  • Introduction of Region Inference for Dynamically-typed Procedural Programming Languages  [in Japanese]

    齋地 崇大 , 前田 敦司

    … Most studies, however, assumes statically-typed languages and few are dealing with dynamically-typed languages and their performance with region-based memory management. … In this presentation, we introduce region inference in dynamically-typed procedural languages by classifying objects into region-inferable and uninferable ones, and discuss the possibility of eliminating the cost related to dynamic memory management such as garbage collection. …

    情報処理学会論文誌プログラミング(PRO) 12(1), 10-10, 2019-01-30

    IPSJ 

  • Latent Words Recurrent Neural Network Language Models for Automatic Speech Recognition

    MASUMURA Ryo , ASAMI Taichi , OBA Takanobu , SAKAUCHI Sumitaka , ITO Akinori

    … The RNN-LMs can capture long-range context information and offer strong performance, and the LW-LMs are robust for out-of-domain tasks based on the latent word space modeling. … This paper also details a parameter inference method and two kinds of implementation methods, an n-gram approximation and a Viterbi approximation, for introducing the LW-LM to ASR. …

    IEICE Transactions on Information and Systems E102.D(12), 2557-2567, 2019

    J-STAGE 

Page Top