Paper Abstract and Keywords |
Presentation |
2022-03-01 14:20
Fast Distortion Pedal Modeling with Fine-Tuning Haruki Shoji, Kento Yoshimoto, Daiki Saka, Hiroki Kuroda, Daichi Kitahara, Kenichiro Tanaka, Akira Hirabayashi (Ritsumeikan Univ.) EA2021-75 SIP2021-102 SP2021-60 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
We propose a fast modeling method for distortion pedals based on deep learning. For modeling many times with different pedals and settings, it is desired to shorten the training time per one model, but simply reducing the amount of training data decreases the modeling accuracy. In this paper, when modeling a target distortion pedal from a small amount of data, we propose to apply fine-tuning where network parameters well-trained for another distortion pedal are used as initial values. Numerical experiments show that the proposed method trains the model of the target distortion pedal very quickly from a small amount of data while maintaining the modeling accuracy. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Distortion Pedal / Deep Learning / WaveNet / Transfer Learning / Fine-Tuning / / / |
Reference Info. |
IEICE Tech. Rep., vol. 121, no. 384, SIP2021-102, pp. 70-75, March 2022. |
Paper # |
SIP2021-102 |
Date of Issue |
2022-02-22 (EA, SIP, SP) |
ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
EA2021-75 SIP2021-102 SP2021-60 |
|