iMTE

논문 제목 : Sanity checks for saliency maps 논문 주소 : arxiv.org/abs/1810.03292 Sanity Checks for Saliency Maps Saliency methods have emerged as a popular tool to highlight features in an input deemed relevant for the prediction of a learned model. Several saliency methods have been proposed, often guided by visual appeal on image data. In this work, we propose an a arxiv.org 주요 내용 : 1) Saliency map은 학..

논문 제목 : Sanity checks for saliency maps 논문 주소 : arxiv.org/abs/1810.03292 Sanity Checks for Saliency Maps Saliency methods have emerged as a popular tool to highlight features in an input deemed relevant for the prediction of a learned model. Several saliency methods have been proposed, often guided by visual appeal on image data. In this work, we propose an a arxiv.org 주요 수식 정리: 0) Definition in..

논문 제목 : SmoothGrad : removing noise by adding noise 논문 주소 : arxiv.org/abs/1706.03825 SmoothGrad: removing noise by adding noise Explaining the output of a deep network remains a challenge. In the case of an image classifier, one type of explanation is to identify pixels that strongly influence the final decision. A starting point for this strategy is the gradient of the class score arxiv.org 주요 ..

논문 제목 : Smooth Grad-CAM++: An Enhanced Inference Level Visualization Technique for Deep Convolutional Neural Network Models 논문 주소 : arxiv.org/abs/1908.01224 Smooth Grad-CAM++: An Enhanced Inference Level Visualization Technique for Deep Convolutional Neural Network Models Gaining insight into how deep convolutional neural network models perform image classification and how to explain their outpu..

논문 제목 : Grad-CAM++: Generalized Gradient-based Visual Explanations for Deep Convolutional Networks 논문 주소 : arxiv.org/pdf/1710.11063.pdf IEEE WACV (2018, ieeexplore.ieee.org/document/8354201)에 나온 논문을 바탕으로 이해하고 내용을 작성한다. arixv에서 나온 버전이 좀 더 extended version임으로 Grad-CAM++에 더 깊은 이해를 위해서는 extended version을 읽는 것을 추천한다. 주요 내용 : 1) Deep models은 "black box"로서 internal function을 이해하는데에는 어려움이 있다. 이를 해결하기 위해..

논문 제목 : Learning deep features for discriminative localization 논문 주소 : openaccess.thecvf.com/content_iccv_2017/html/Selvaraju_Grad-CAM_Visual_Explanations_ICCV_2017_paper.html ICCV 2017 Open Access Repository Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, Dhruv Batra; Pr..

논문 제목 : Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization 논문 주소 : openaccess.thecvf.com/content_iccv_2017/html/Selvaraju_Grad-CAM_Visual_Explanations_ICCV_2017_paper.html ICCV 2017 Open Access Repository Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, De..

activation functions In [1]: import matplotlib.pyplot as plt import numpy as np Step function¶ In [2]: def step(x): return 1*(x>0) In [3]: inputs = np.arange(-5,5,0.01) outputs = step(inputs) plt.figure(figsize=(8,5)) plt.plot(inputs,outputs,label='Step function') plt.hlines(0,-5,5) plt.vlines(0,0,1) plt.xlabel('input',fontsize=24) plt.ylabel('output',fontsize=24) plt.grid(alpha=0.3) plt.title("..

Meta-learning with Implicit Gradients [1] https://papers.nips.cc/paper/2019/hash/072b030ba126b2f4b2374f342be9ed44-Abstract.html IntroductionMeta-learning의 frame에서 bi-level optimization procedure는 다음으로 나누어진다.1) inner optimization : 주어진 task에 base learner가 학습하는 과정2) outer optimization : 여러 tasks 들에서 meta learner가 학습하는 과정MAML, DAML, Reptile 등의 방법이 optimization-based methods에 속한다. (Hands-on one-shot..

다시 복귀 할 예정입니다.... 지금까지 논문 작업 하고 있어서 블로그에 집중하지 못했는데, 이제 다시 블로그하면서 공부하려고 합니다.주로 GAN과 meta-learning or one-shot learning을 다룰 예정입니다.감사합니다.