• Journal of the Korea Institute of Military Science and Technology
• /
• v.27 no.2
• /
• pp.256-264
• /
• 2024

In the last 10 years, AI has made rapid progress, and image classification, in particular, are showing excellent performance based on deep learning. Nevertheless, due to the nature of deep learning represented by a black box, it is difficult to actually use it in critical decision-making situations such as national defense, autonomous driving, medical care, and finance due to the lack of explainability of judgement results. In order to overcome these limitations, in this study, a model description algorithm capable of local interpretation was applied to the inception network-derived AI to analyze what grounds they made when classifying national defense data. Specifically, we conduct a comparative analysis of explainability based on confidence values by performing LIME analysis from the Inception v2_resnet model and verify the similarity between human interpretations and LIME explanations. Furthermore, by comparing the LIME explanation results through the Top1 output results for Inception v3, Inception v2_resnet, and Xception models, we confirm the feasibility of comparing the efficiency and availability of deep learning networks using XAI.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2017.11a
• /
• pp.257-258
• /
• 2017

머신러닝 기법을 다양한 분야에 사용되는 연구가 한창이다. 본 논문에서는 악성 코드의 분류 시스템에 머신러닝 기법을 적용하였다. 악성 코드 파일을 적당한 크기로 이미지화하여 텐서 플로우의 인셉션 V3에 적용하였다. 실험 결과, 이미지의 사이즈 조정과 파라미터 조정을 통해 매우 만족할 만한 수준으로 악성 코드를 잘 분류함을 확인할 수 있었다.

• Journal of the Institute of Convergence Signal Processing
• /
• v.20 no.3
• /
• pp.132-137
• /
• 2019

The amount of data generated from medical images is increasingly exceeding the limits of professional visual analysis, and the need for automated medical image analysis is increasing. For this reason, this study evaluated the classification and accuracy according to the presence or absence of tumor using Inception V3 deep learning model, using MRI medical images showing normal and tumor findings. As a result, the accuracy of the deep learning model was 90% for the training data set and 86% for the validation data set. The loss rate was 0.56 for the training data set and 1.28 for the validation data set. In future studies, it is necessary to secure the data of publicly available medical images to improve the performance of the deep learning model and to ensure the reliability of the evaluation, and to implement modeling by improving the accuracy of labeling through labeling classification.