Learning Human Perceptual Hysteresis via Deep CNN-RNN Networks
International Journal of High School Research · 30.04.2026 · 8 (6) · 69-73
DOI: 10.36838/ijhsr86.69 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of High School Research, online erschienen am 30.04.2026. Verfasst von Jason Hwang.
Abstract (Originalsprache)
For many years, neuroscientists have been interested in modeling human perceptual hysteresis.However, previous models were hard to tune and could not generalize across different datasets.To address this, we propose a supervised convolutional neural network -recurrent neural network (CNN-RNN) that learns hysteresis directly from data instead of separate simulation models.To control the hysteresis effect, we also introduce a regularization term that penalizes changes in low-confidence classes.We show that our CNN-RNN model successfully mimics human perceptual hysteresis across a large number of examples.Our model is expected to decrease radiologists' hysteresis on disease screens by finding a sequence of images that has no perceptual hysteresis.
Zitieren
APA
Hwang, J. (2026). Learning Human Perceptual Hysteresis via Deep CNN-RNN Networks. International Journal of High School Research, 8(6), 69-73. https://doi.org/10.36838/ijhsr86.69
Vancouver
Hwang J. Learning Human Perceptual Hysteresis via Deep CNN-RNN Networks. International Journal of High School Research. 2026;8(6):69-73. doi:10.36838/ijhsr86.69
BibTeX
@article{Hwang2026,
author = {Hwang, Jason},
title = {Learning Human Perceptual Hysteresis via Deep CNN-RNN Networks},
journal = {International Journal of High School Research},
year = {2026},
volume = {8},
number = {6},
pages = {69--73},
doi = {10.36838/ijhsr86.69}
}
EUVIMED