JSAI2020

Presentation information

Interactive Session

[4Rin1] Interactive 2

Fri. Jun 12, 2020 9:00 AM - 10:40 AM Room R01 (jsai2020online-2-33)

[4Rin1-24] A Study on Attention Mechanism in Deep Learning

〇Rina Hirota1, Ichiro Kobayashi1 (1.Ochanomizu University)

Keywords:Attention Mechanism, Image captioning, Explainability

Recently, artificial intelligence has demonstrated remarkable performance in many tasks, especially with the advance of deep learning techniques. For instance, deep learning models are capable of recognizing images with high accuracy. However, due to their black-box nature, the way they perform decisions is still poorly understood. Fields such as the medical field, for example, require a high level of accountability, and thus transparency. Therefore, we need to be able to explain machine decisions and justify their reliability. In this study, we explore the attention mechanism of neural networks and propose a model that not only recognizes images but also outputs their textual explanation as well as attention visualization. Our final goal is to develop an explainable AI model that can be more reliable when performing real-world tasks.

Authentication for paper PDF access

A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.

Password