H
Hayaru Shouno
Researcher at University of Electro-Communications
Publications - 80
Citations - 615
Hayaru Shouno is an academic researcher from University of Electro-Communications. The author has contributed to research in topics: Deep learning & Computer science. The author has an hindex of 9, co-authored 69 publications receiving 445 citations. Previous affiliations of Hayaru Shouno include Osaka University & Yamaguchi University.
Papers
More filters
Proceedings ArticleDOI
Analysis of function of rectified linear unit used in deep learning
TL;DR: A rectified linear unit (ReLU) is proposed to speed up the learning convergence of the deep learning using a using simpler network called the soft-committee machine and the reasons for the speedup are clarified.
Book ChapterDOI
Analysis of Dropout Learning Regarded as Ensemble Learning
TL;DR: It is found that the process of combining the neglected hidden units with the learned network can be regarded as ensemble learning, so dropout learning is analyzed from this point of view.
Journal ArticleDOI
Accuracy of the Bethe approximation for hyperparameter estimation in probabilistic image processing
TL;DR: This paper investigates the accuracy of statistical-mechanical approximations for the estimation of hyperparameters from observable data in probabilistic image processing, which is based on Bayesian statistics and maximum likelihood estimation.
Book ChapterDOI
Analysis of dropout learning regarded as ensemble learning
TL;DR: In this article, the authors proposed Dropout Learning, which neglects some inputs and hidden units in the learning process with a probability, p, and then, the neglected input and hidden unit are combined with the learned network to express the final output.
Proceedings ArticleDOI
Training neocognitron to recognize handwritten digits in the real world
TL;DR: Using a large-scale real-world database, it is shown that a neocognitron trained by unsupervised learning with a winner-take-all process can recognize handwritten digits with a recognition rate higher than 97%.