DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | 최재식 | - |
dc.contributor.author | Park, Bumjin | - |
dc.contributor.author | 박범진 | - |
dc.date.accessioned | 2024-07-25T19:30:47Z | - |
dc.date.available | 2024-07-25T19:30:47Z | - |
dc.date.issued | 2023 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=1045733&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/320545 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 김재철AI대학원, 2023.8,[iii, 22 p. :] | - |
dc.description.abstract | Gradient signals have been widely used for the interpretability of a convolution neural network. To explain the decision of a single input, all channels in a layer contribute to gradient propagation. We hypothesize that not all channels are required to explain a single input and that channel pruning can improve the reliability of a saliency map. To test this hypothesis, we propose what is termed the partitioned channel gradient (ParchGrad), which partitions channels into two sets and modifies the gradient signals so that the ratio of the gradient magnitudes is manually controllable. In addition, we propose simple channel partitioning methods to prune channels for ParchGrad. We empirically show that \ours, combined with several saliency methods, results in a more reliable saliency map than the original gradient signal. Also, we found that (1) only a few channels (~10%) are required to explain a single input and (2) that the optimal pruning layers are different for each class label. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | 설명가능인공지능▼a딥러닝▼a채널제거▼a입력기여도 | - |
dc.subject | Explainable AI▼aDeep learning▼aChannel pruning▼aInput attribution | - |
dc.title | Partitioned channel gradient for reliable saliency map in image classification | - |
dc.title.alternative | 합성곱 신경망의 분할된 채널을 활용한 입력 기여도 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :김재철AI대학원, | - |
dc.contributor.alternativeauthor | Choi, Jaesik | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.