SlideShare a Scribd company logo
2021년 5월 23일
딥러닝 논문읽기 모임
이미지 처리팀 : 김병현 안종식 이찬혁 홍은기
NBDT : Neural-backed Decision Tree
Wan, Alvin et al. (UC Berkely)
ICLR 2021
Contents
Introduction 01
03
02
04
05
Experiments
Related Works
Method
Conclusion
Introduction
01
01. Introduction
1. Necessity of Research
[ PCB Inspection System ]
Open Defect Image Traditional Image Processing
Post-Processing
Thresholding
Find Condition
Open Defect !
01. Introduction
1. Necessity of Research
[ PCB Inspection System ]
Open Defect Image Deep Learning Based
Model Probability of “open” : 99%
Probability of “Short” : 1%
Probability of “Particle” : 5%
Deep Learning is Black Box !
01. Introduction
1. Necessity of Research
01. Introduction
2. Interpretability
1) Manipulating and Measuring Model Interpretability, Microsoft, 2021
Able to debug Model or Dataset Improving human trust
Related Works
02
02. Related Works
1. Saliency Map
1) Grad-CAM : Visual Explanations from Deep Networks via Gradient-based Localization
Input Image Saliency map Guided Backpropagation
02. Related Works
1. Saliency Map
1) Grad-CAM : Visual Explanations from Deep Networks via Gradient-based Localization
- Output을 도출할 때, Input의 어느 부분이 영향을 얼마나 주는가 ?
검은목논병아리 귀뿔논병아리
Not enough to explain “Why?”
02. Related Works
2. Sequential Decision Processes
1) Decision Tree
- Rule Base Model과 같이 의사결정 단계를 작게 쪼게 확인 가능
Oblique Decision Tree
Decision Tree
Q & A
Method
03
03. Method
1. Architecture
- Classification Head의 마지막 FCN Layer 뒤에 Oblique Decision Tree가 추가적으로 Attach
Feature Extractor
Classification Head
(FCN)
Can Apply to Any Classification Models!
Induced Hierarchies
03. Method
2. Build Induced Hierarchies
1) Pre-train Model
- NBDTs는 Pre-train Model이 필요 : Pretrain Model의 마지막 FCN Layer Weight를 이용해 Hierarchies 구성
Build
Hierarchies
Inference
03. Method
2. Build Induced Hierarchies
2) Agglomerative Clustering
- build Hierarchies
03. Method
2. Build Induced Hierarchies
2) Agglomerative Clustering
- build Hierarchies
k
Weight Space
Clustering Repeat
03. Method
2. Build Induced Hierarchies
2) Agglomerative Clustering
- build Hierarchies
Problem 1. Problem 2.
?
?
03. Method
2. Build Induced Hierarchies
3) WordNet : 단어들을 계층적으로 구조화 해 놓은 것
- Clustering 된 두 Class들의 공통 조상을 WordNet에서 찾음
[ WordNet ]
CIFAR-100
03. Method
2. Build Induced Hierarchies
4) Set Weight of Induced Hierarchies
Step A,B) Load Pretrain Model’s Weight & Allocate Weight to each Leaf Nodes
Step C) Set Parent Vector : Average of Child Node
Step D) Repeat Until Set all node’s weight
𝑣𝑒𝑐𝑡𝑜𝑟 𝑛1
d
k
Average
03. Method
3. Inference
1) Soft Decision Tree vs Hard Decision Tree
Inner
Product
- Probability of class K - Final Prediction
“분기” 개념보다
“Path” 개념에 가깝다!
Softmax
03. Method
4. Loss
1) Tree Supervision Loss
03. Method
4. Loss
1) Tree Supervision Loss
Feature Extractor
Classification Head
(FCN)
Induced Hierarchies
0  0.5 1  0
Q & A
EXPERIMENTS
04
04. Experiments
1. Accuracy
1) Small-scale Datasets
04. Experiments
1. Accuracy
2) Large-scale Datasets
04. Experiments
2. Ablation Study
1) Hierarchies
04. Experiments
2. Ablation Study
2) Losses
04. Experiments
3. Interpretability : Survey
1) 방법 : Saliency Map과 NBDT 설명이 주어졌을 때, 어느 것이 더 옳게 찾는지 Survey
2) 종류
a) 주어진 설명을 통해서 Model의 오탐지를 찾을 수 있는가?
- 결과 : 600개의 응답 중 Saliency Map은 87개 정답 / NBDT는 237개 정답
b) 어떤 모델의 판단을 더 신뢰하는가?
- 결과 : 374개의 응답 중 65.9%가 NBDT를 더 신뢰함
04. Experiments
3. Interpretability : Survey
c) 판단이 어려운 상황에서 모델의 예측을 신뢰할 수 있는가?
- 매우 Blur한 이미지와 각 모델이 예측한 결과를 제시
> 이때, Prediction의 결과는 다름 (NBDT가 정답 : Saliency Map가 정답 : 둘 다 틀림 = 3 : 3 : 4)
- 결과 : 600개의 응답 중 167개 Saliency Map의 결과에 동의, 312개 NBDT의 결과에 동의, 나머지는 둘다 반대
> 600개의 결과 중 255개의 응답이 정답을 맞춤
04. Experiments
3. Interpretability : Analysis
2) Model이 헷갈리는 이미지를 찾는데 사용될 수 있는 가?
- 헷갈리는 문제에서, 확실한 부분은 높은 Probability를, 헷갈리는 부분에서는 낮은 Probability를 보인다
- 위를 이용하여 “Path Entropy”를 계산하여 헷갈리는 Image를 구분해낼 수 있다.
04. Experiments
3. Interpretability : Analysis
3) Model이 헷갈리는 이미지를 찾는데 사용될 수 있는 가?
- ImageNet 내의 Multi Target Image를 분류하는데 사용 가능
Conclusion
05
05. Conclusion
1. Contribution
1) Accuracy를 유지하며 Interpretability를 향상
2) Grad-CAM 기법에 비해 Interpretability가 높음
3) 모든 Model에 적용 가능
2. Discussion
1) Good
- Survey 기반으로 Human Trust를 수치화 한 것은 인상적
- ImageNet의 잘못된 Image를 분리해 낼 수 있다 : Sub-Result
2) Bad
- 논문 전반적으로 혼동이 올 수 있는 언어 사용 / 설명 Skip이 많음
Q & A
THANK YOU
for Watching

More Related Content

What's hot

MixMatch: A Holistic Approach to Semi- Supervised Learning
MixMatch: A Holistic Approach to Semi- Supervised LearningMixMatch: A Holistic Approach to Semi- Supervised Learning
MixMatch: A Holistic Approach to Semi- Supervised Learning
harmonylab
 
[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ
[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ
[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ
Deep Learning JP
 
車載カメラの映像から歩行者に関わる危険を予測する技術
車載カメラの映像から歩行者に関わる危険を予測する技術車載カメラの映像から歩行者に関わる危険を予測する技術
車載カメラの映像から歩行者に関わる危険を予測する技術
Takuya Minagawa
 
SSII2014 チュートリアル資料
SSII2014 チュートリアル資料SSII2014 チュートリアル資料
SSII2014 チュートリアル資料Masayuki Tanaka
 
SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法
SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法
SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法
SSII
 
自然言語処理における深層学習を用いた予測の不確実性 - Predictive Uncertainty in NLP -
自然言語処理における深層学習を用いた予測の不確実性  - Predictive Uncertainty in NLP -自然言語処理における深層学習を用いた予測の不確実性  - Predictive Uncertainty in NLP -
自然言語処理における深層学習を用いた予測の不確実性 - Predictive Uncertainty in NLP -
tmtm otm
 
モデル高速化百選
モデル高速化百選モデル高速化百選
モデル高速化百選
Yusuke Uchida
 
2014.02.20_5章ニューラルネットワーク
2014.02.20_5章ニューラルネットワーク2014.02.20_5章ニューラルネットワーク
2014.02.20_5章ニューラルネットワークTakeshi Sakaki
 
ICML 2021 Workshop 深層学習の不確実性について
ICML 2021 Workshop 深層学習の不確実性についてICML 2021 Workshop 深層学習の不確実性について
ICML 2021 Workshop 深層学習の不確実性について
tmtm otm
 
[DL輪読会]Pay Attention to MLPs (gMLP)
[DL輪読会]Pay Attention to MLPs	(gMLP)[DL輪読会]Pay Attention to MLPs	(gMLP)
[DL輪読会]Pay Attention to MLPs (gMLP)
Deep Learning JP
 
【DL輪読会】Transformers are Sample Efficient World Models
【DL輪読会】Transformers are Sample Efficient World Models【DL輪読会】Transformers are Sample Efficient World Models
【DL輪読会】Transformers are Sample Efficient World Models
Deep Learning JP
 
読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法
読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法
読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法
健児 青木
 
머피's 머신러닝, Mixture model and EM algorithm
머피's 머신러닝, Mixture model and EM algorithm머피's 머신러닝, Mixture model and EM algorithm
머피's 머신러닝, Mixture model and EM algorithm
Jungkyu Lee
 
Efficient estimation of word representations in vector space
Efficient estimation of word representations in vector spaceEfficient estimation of word representations in vector space
Efficient estimation of word representations in vector space
tetsuo ishigaki
 
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
Deep Learning JP
 
[DL輪読会]Are Sixteen Heads Really Better than One?
[DL輪読会]Are Sixteen Heads Really Better than One?[DL輪読会]Are Sixteen Heads Really Better than One?
[DL輪読会]Are Sixteen Heads Really Better than One?
Deep Learning JP
 
[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...
[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...
[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...
Deep Learning JP
 
Cosine Based Softmax による Metric Learning が上手くいく理由
Cosine Based Softmax による Metric Learning が上手くいく理由Cosine Based Softmax による Metric Learning が上手くいく理由
Cosine Based Softmax による Metric Learning が上手くいく理由
tancoro
 
【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?
【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?
【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?
Deep Learning JP
 

What's hot (20)

MixMatch: A Holistic Approach to Semi- Supervised Learning
MixMatch: A Holistic Approach to Semi- Supervised LearningMixMatch: A Holistic Approach to Semi- Supervised Learning
MixMatch: A Holistic Approach to Semi- Supervised Learning
 
[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ
[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ
[DL輪読会]物理学による帰納バイアスを組み込んだダイナミクスモデル作成に関する論文まとめ
 
車載カメラの映像から歩行者に関わる危険を予測する技術
車載カメラの映像から歩行者に関わる危険を予測する技術車載カメラの映像から歩行者に関わる危険を予測する技術
車載カメラの映像から歩行者に関わる危険を予測する技術
 
SSII2014 チュートリアル資料
SSII2014 チュートリアル資料SSII2014 チュートリアル資料
SSII2014 チュートリアル資料
 
SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法
SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法
SSII2021 [OS2-01] 転移学習の基礎:異なるタスクの知識を利用するための機械学習の方法
 
自然言語処理における深層学習を用いた予測の不確実性 - Predictive Uncertainty in NLP -
自然言語処理における深層学習を用いた予測の不確実性  - Predictive Uncertainty in NLP -自然言語処理における深層学習を用いた予測の不確実性  - Predictive Uncertainty in NLP -
自然言語処理における深層学習を用いた予測の不確実性 - Predictive Uncertainty in NLP -
 
モデル高速化百選
モデル高速化百選モデル高速化百選
モデル高速化百選
 
2014.02.20_5章ニューラルネットワーク
2014.02.20_5章ニューラルネットワーク2014.02.20_5章ニューラルネットワーク
2014.02.20_5章ニューラルネットワーク
 
Prml14 5
Prml14 5Prml14 5
Prml14 5
 
ICML 2021 Workshop 深層学習の不確実性について
ICML 2021 Workshop 深層学習の不確実性についてICML 2021 Workshop 深層学習の不確実性について
ICML 2021 Workshop 深層学習の不確実性について
 
[DL輪読会]Pay Attention to MLPs (gMLP)
[DL輪読会]Pay Attention to MLPs	(gMLP)[DL輪読会]Pay Attention to MLPs	(gMLP)
[DL輪読会]Pay Attention to MLPs (gMLP)
 
【DL輪読会】Transformers are Sample Efficient World Models
【DL輪読会】Transformers are Sample Efficient World Models【DL輪読会】Transformers are Sample Efficient World Models
【DL輪読会】Transformers are Sample Efficient World Models
 
読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法
読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法
読書会 「トピックモデルによる統計的潜在意味解析」 第2回 3.2節 サンプリング近似法
 
머피's 머신러닝, Mixture model and EM algorithm
머피's 머신러닝, Mixture model and EM algorithm머피's 머신러닝, Mixture model and EM algorithm
머피's 머신러닝, Mixture model and EM algorithm
 
Efficient estimation of word representations in vector space
Efficient estimation of word representations in vector spaceEfficient estimation of word representations in vector space
Efficient estimation of word representations in vector space
 
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
【DL輪読会】Contrastive Learning as Goal-Conditioned Reinforcement Learning
 
[DL輪読会]Are Sixteen Heads Really Better than One?
[DL輪読会]Are Sixteen Heads Really Better than One?[DL輪読会]Are Sixteen Heads Really Better than One?
[DL輪読会]Are Sixteen Heads Really Better than One?
 
[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...
[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...
[DL輪読会]RobustNet: Improving Domain Generalization in Urban- Scene Segmentatio...
 
Cosine Based Softmax による Metric Learning が上手くいく理由
Cosine Based Softmax による Metric Learning が上手くいく理由Cosine Based Softmax による Metric Learning が上手くいく理由
Cosine Based Softmax による Metric Learning が上手くいく理由
 
【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?
【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?
【DL輪読会】Is Conditional Generative Modeling All You Need For Decision-Making?
 

Similar to NBDT : Neural-backed Decision Tree 2021 ICLR

Fin_whales
Fin_whalesFin_whales
Fin_whales
Elijah Willie
 
PPT - Deep and Confident Prediction For Time Series at Uber
PPT - Deep and Confident Prediction For Time Series at UberPPT - Deep and Confident Prediction For Time Series at Uber
PPT - Deep and Confident Prediction For Time Series at Uber
Jisang Yoon
 
Literature Survey on Image Deblurring Techniques
Literature Survey on Image Deblurring TechniquesLiterature Survey on Image Deblurring Techniques
Literature Survey on Image Deblurring Techniques
Editor IJCATR
 
Deep Learning for Computer Vision - PyconDE 2017
Deep Learning for Computer Vision - PyconDE 2017Deep Learning for Computer Vision - PyconDE 2017
Deep Learning for Computer Vision - PyconDE 2017
Alex Conway
 
Learning where to look: focus and attention in deep vision
Learning where to look: focus and attention in deep visionLearning where to look: focus and attention in deep vision
Learning where to look: focus and attention in deep vision
Universitat Politècnica de Catalunya
 
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
IRJET Journal
 
researchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdf
researchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdfresearchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdf
researchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdf
AvijitChaudhuri3
 
Types of Machine Learnig Algorithms(CART, ID3)
Types of Machine Learnig Algorithms(CART, ID3)Types of Machine Learnig Algorithms(CART, ID3)
Types of Machine Learnig Algorithms(CART, ID3)
Fatimakhan325
 
Cluster Analysis : Assignment & Update
Cluster Analysis : Assignment & UpdateCluster Analysis : Assignment & Update
Cluster Analysis : Assignment & Update
Billy Yang
 
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...
CSCJournals
 
NIPS2007: deep belief nets
NIPS2007: deep belief netsNIPS2007: deep belief nets
NIPS2007: deep belief nets
zukun
 
세미나 20170929
세미나 20170929세미나 20170929
세미나 20170929
Lee Gyeong Hoon
 
Report face recognition : ArganRecogn
Report face recognition :  ArganRecognReport face recognition :  ArganRecogn
Report face recognition : ArganRecogn
Ilyas CHAOUA
 
Final Paper 2
Final Paper 2Final Paper 2
Final Paper 2
Elizabeth Koshelev
 
Deep learning in Computer Vision
Deep learning in Computer VisionDeep learning in Computer Vision
Deep learning in Computer Vision
David Dao
 
Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation
Sneha Ravikumar
 
Declarative data analysis
Declarative data analysisDeclarative data analysis
Declarative data analysis
South West Data Meetup
 
neuralAC
neuralACneuralAC
neuralAC
Dr Rupesh Shet
 
Predicting rainfall using ensemble of ensembles
Predicting rainfall using ensemble of ensemblesPredicting rainfall using ensemble of ensembles
Predicting rainfall using ensemble of ensembles
Varad Meru
 
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Chris Rackauckas
 

Similar to NBDT : Neural-backed Decision Tree 2021 ICLR (20)

Fin_whales
Fin_whalesFin_whales
Fin_whales
 
PPT - Deep and Confident Prediction For Time Series at Uber
PPT - Deep and Confident Prediction For Time Series at UberPPT - Deep and Confident Prediction For Time Series at Uber
PPT - Deep and Confident Prediction For Time Series at Uber
 
Literature Survey on Image Deblurring Techniques
Literature Survey on Image Deblurring TechniquesLiterature Survey on Image Deblurring Techniques
Literature Survey on Image Deblurring Techniques
 
Deep Learning for Computer Vision - PyconDE 2017
Deep Learning for Computer Vision - PyconDE 2017Deep Learning for Computer Vision - PyconDE 2017
Deep Learning for Computer Vision - PyconDE 2017
 
Learning where to look: focus and attention in deep vision
Learning where to look: focus and attention in deep visionLearning where to look: focus and attention in deep vision
Learning where to look: focus and attention in deep vision
 
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
An Approach for Image Deblurring: Based on Sparse Representation and Regulari...
 
researchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdf
researchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdfresearchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdf
researchpaper_2023_Skin_Csdbjsjvnvsdnfvancer.pdf
 
Types of Machine Learnig Algorithms(CART, ID3)
Types of Machine Learnig Algorithms(CART, ID3)Types of Machine Learnig Algorithms(CART, ID3)
Types of Machine Learnig Algorithms(CART, ID3)
 
Cluster Analysis : Assignment & Update
Cluster Analysis : Assignment & UpdateCluster Analysis : Assignment & Update
Cluster Analysis : Assignment & Update
 
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...
Shallow vs. Deep Image Representations: A Comparative Study with Enhancements...
 
NIPS2007: deep belief nets
NIPS2007: deep belief netsNIPS2007: deep belief nets
NIPS2007: deep belief nets
 
세미나 20170929
세미나 20170929세미나 20170929
세미나 20170929
 
Report face recognition : ArganRecogn
Report face recognition :  ArganRecognReport face recognition :  ArganRecogn
Report face recognition : ArganRecogn
 
Final Paper 2
Final Paper 2Final Paper 2
Final Paper 2
 
Deep learning in Computer Vision
Deep learning in Computer VisionDeep learning in Computer Vision
Deep learning in Computer Vision
 
Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation
 
Declarative data analysis
Declarative data analysisDeclarative data analysis
Declarative data analysis
 
neuralAC
neuralACneuralAC
neuralAC
 
Predicting rainfall using ensemble of ensembles
Predicting rainfall using ensemble of ensemblesPredicting rainfall using ensemble of ensembles
Predicting rainfall using ensemble of ensembles
 
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
Generalizing Scientific Machine Learning and Differentiable Simulation Beyond...
 

More from taeseon ryu

VoxelNet
VoxelNetVoxelNet
VoxelNet
taeseon ryu
 
OpineSum Entailment-based self-training for abstractive opinion summarization...
OpineSum Entailment-based self-training for abstractive opinion summarization...OpineSum Entailment-based self-training for abstractive opinion summarization...
OpineSum Entailment-based self-training for abstractive opinion summarization...
taeseon ryu
 
3D Gaussian Splatting
3D Gaussian Splatting3D Gaussian Splatting
3D Gaussian Splatting
taeseon ryu
 
JetsonTX2 Python
 JetsonTX2 Python  JetsonTX2 Python
JetsonTX2 Python
taeseon ryu
 
Hyperbolic Image Embedding.pptx
Hyperbolic  Image Embedding.pptxHyperbolic  Image Embedding.pptx
Hyperbolic Image Embedding.pptx
taeseon ryu
 
MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정
MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정
MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정
taeseon ryu
 
LLaMA Open and Efficient Foundation Language Models - 230528.pdf
LLaMA Open and Efficient Foundation Language Models - 230528.pdfLLaMA Open and Efficient Foundation Language Models - 230528.pdf
LLaMA Open and Efficient Foundation Language Models - 230528.pdf
taeseon ryu
 
YOLO V6
YOLO V6YOLO V6
YOLO V6
taeseon ryu
 
Dataset Distillation by Matching Training Trajectories
Dataset Distillation by Matching Training Trajectories Dataset Distillation by Matching Training Trajectories
Dataset Distillation by Matching Training Trajectories
taeseon ryu
 
RL_UpsideDown
RL_UpsideDownRL_UpsideDown
RL_UpsideDown
taeseon ryu
 
Packed Levitated Marker for Entity and Relation Extraction
Packed Levitated Marker for Entity and Relation ExtractionPacked Levitated Marker for Entity and Relation Extraction
Packed Levitated Marker for Entity and Relation Extraction
taeseon ryu
 
MOReL: Model-Based Offline Reinforcement Learning
MOReL: Model-Based Offline Reinforcement LearningMOReL: Model-Based Offline Reinforcement Learning
MOReL: Model-Based Offline Reinforcement Learning
taeseon ryu
 
Scaling Instruction-Finetuned Language Models
Scaling Instruction-Finetuned Language ModelsScaling Instruction-Finetuned Language Models
Scaling Instruction-Finetuned Language Models
taeseon ryu
 
Visual prompt tuning
Visual prompt tuningVisual prompt tuning
Visual prompt tuning
taeseon ryu
 
mPLUG
mPLUGmPLUG
variBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdf
variBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdfvariBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdf
variBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdf
taeseon ryu
 
Reinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdf
Reinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdfReinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdf
Reinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdf
taeseon ryu
 
The Forward-Forward Algorithm
The Forward-Forward AlgorithmThe Forward-Forward Algorithm
The Forward-Forward Algorithm
taeseon ryu
 
Towards Robust and Reproducible Active Learning using Neural Networks
Towards Robust and Reproducible Active Learning using Neural NetworksTowards Robust and Reproducible Active Learning using Neural Networks
Towards Robust and Reproducible Active Learning using Neural Networks
taeseon ryu
 
BRIO: Bringing Order to Abstractive Summarization
BRIO: Bringing Order to Abstractive SummarizationBRIO: Bringing Order to Abstractive Summarization
BRIO: Bringing Order to Abstractive Summarization
taeseon ryu
 

More from taeseon ryu (20)

VoxelNet
VoxelNetVoxelNet
VoxelNet
 
OpineSum Entailment-based self-training for abstractive opinion summarization...
OpineSum Entailment-based self-training for abstractive opinion summarization...OpineSum Entailment-based self-training for abstractive opinion summarization...
OpineSum Entailment-based self-training for abstractive opinion summarization...
 
3D Gaussian Splatting
3D Gaussian Splatting3D Gaussian Splatting
3D Gaussian Splatting
 
JetsonTX2 Python
 JetsonTX2 Python  JetsonTX2 Python
JetsonTX2 Python
 
Hyperbolic Image Embedding.pptx
Hyperbolic  Image Embedding.pptxHyperbolic  Image Embedding.pptx
Hyperbolic Image Embedding.pptx
 
MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정
MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정
MCSE_Multimodal Contrastive Learning of Sentence Embeddings_변현정
 
LLaMA Open and Efficient Foundation Language Models - 230528.pdf
LLaMA Open and Efficient Foundation Language Models - 230528.pdfLLaMA Open and Efficient Foundation Language Models - 230528.pdf
LLaMA Open and Efficient Foundation Language Models - 230528.pdf
 
YOLO V6
YOLO V6YOLO V6
YOLO V6
 
Dataset Distillation by Matching Training Trajectories
Dataset Distillation by Matching Training Trajectories Dataset Distillation by Matching Training Trajectories
Dataset Distillation by Matching Training Trajectories
 
RL_UpsideDown
RL_UpsideDownRL_UpsideDown
RL_UpsideDown
 
Packed Levitated Marker for Entity and Relation Extraction
Packed Levitated Marker for Entity and Relation ExtractionPacked Levitated Marker for Entity and Relation Extraction
Packed Levitated Marker for Entity and Relation Extraction
 
MOReL: Model-Based Offline Reinforcement Learning
MOReL: Model-Based Offline Reinforcement LearningMOReL: Model-Based Offline Reinforcement Learning
MOReL: Model-Based Offline Reinforcement Learning
 
Scaling Instruction-Finetuned Language Models
Scaling Instruction-Finetuned Language ModelsScaling Instruction-Finetuned Language Models
Scaling Instruction-Finetuned Language Models
 
Visual prompt tuning
Visual prompt tuningVisual prompt tuning
Visual prompt tuning
 
mPLUG
mPLUGmPLUG
mPLUG
 
variBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdf
variBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdfvariBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdf
variBAD, A Very Good Method for Bayes-Adaptive Deep RL via Meta-Learning.pdf
 
Reinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdf
Reinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdfReinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdf
Reinforced Genetic Algorithm Learning For Optimizing Computation Graphs.pdf
 
The Forward-Forward Algorithm
The Forward-Forward AlgorithmThe Forward-Forward Algorithm
The Forward-Forward Algorithm
 
Towards Robust and Reproducible Active Learning using Neural Networks
Towards Robust and Reproducible Active Learning using Neural NetworksTowards Robust and Reproducible Active Learning using Neural Networks
Towards Robust and Reproducible Active Learning using Neural Networks
 
BRIO: Bringing Order to Abstractive Summarization
BRIO: Bringing Order to Abstractive SummarizationBRIO: Bringing Order to Abstractive Summarization
BRIO: Bringing Order to Abstractive Summarization
 

Recently uploaded

Applied Science: Thermodynamics, Laws & Methodology.pdf
Applied Science: Thermodynamics, Laws & Methodology.pdfApplied Science: Thermodynamics, Laws & Methodology.pdf
Applied Science: Thermodynamics, Laws & Methodology.pdf
University of Hertfordshire
 
Medical Orthopedic PowerPoint Templates.pptx
Medical Orthopedic PowerPoint Templates.pptxMedical Orthopedic PowerPoint Templates.pptx
Medical Orthopedic PowerPoint Templates.pptx
terusbelajar5
 
在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样
在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样
在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样
vluwdy49
 
Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...
Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...
Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...
Travis Hills MN
 
Eukaryotic Transcription Presentation.pptx
Eukaryotic Transcription Presentation.pptxEukaryotic Transcription Presentation.pptx
Eukaryotic Transcription Presentation.pptx
RitabrataSarkar3
 
mô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốt
mô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốtmô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốt
mô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốt
HongcNguyn6
 
Deep Software Variability and Frictionless Reproducibility
Deep Software Variability and Frictionless ReproducibilityDeep Software Variability and Frictionless Reproducibility
Deep Software Variability and Frictionless Reproducibility
University of Rennes, INSA Rennes, Inria/IRISA, CNRS
 
20240520 Planning a Circuit Simulator in JavaScript.pptx
20240520 Planning a Circuit Simulator in JavaScript.pptx20240520 Planning a Circuit Simulator in JavaScript.pptx
20240520 Planning a Circuit Simulator in JavaScript.pptx
Sharon Liu
 
Compexometric titration/Chelatorphy titration/chelating titration
Compexometric titration/Chelatorphy titration/chelating titrationCompexometric titration/Chelatorphy titration/chelating titration
Compexometric titration/Chelatorphy titration/chelating titration
Vandana Devesh Sharma
 
Immersive Learning That Works: Research Grounding and Paths Forward
Immersive Learning That Works: Research Grounding and Paths ForwardImmersive Learning That Works: Research Grounding and Paths Forward
Immersive Learning That Works: Research Grounding and Paths Forward
Leonel Morgado
 
Cytokines and their role in immune regulation.pptx
Cytokines and their role in immune regulation.pptxCytokines and their role in immune regulation.pptx
Cytokines and their role in immune regulation.pptx
Hitesh Sikarwar
 
Randomised Optimisation Algorithms in DAPHNE
Randomised Optimisation Algorithms in DAPHNERandomised Optimisation Algorithms in DAPHNE
Randomised Optimisation Algorithms in DAPHNE
University of Maribor
 
Equivariant neural networks and representation theory
Equivariant neural networks and representation theoryEquivariant neural networks and representation theory
Equivariant neural networks and representation theory
Daniel Tubbenhauer
 
Phenomics assisted breeding in crop improvement
Phenomics assisted breeding in crop improvementPhenomics assisted breeding in crop improvement
Phenomics assisted breeding in crop improvement
IshaGoswami9
 
aziz sancar nobel prize winner: from mardin to nobel
aziz sancar nobel prize winner: from mardin to nobelaziz sancar nobel prize winner: from mardin to nobel
aziz sancar nobel prize winner: from mardin to nobel
İsa Badur
 
Describing and Interpreting an Immersive Learning Case with the Immersion Cub...
Describing and Interpreting an Immersive Learning Case with the Immersion Cub...Describing and Interpreting an Immersive Learning Case with the Immersion Cub...
Describing and Interpreting an Immersive Learning Case with the Immersion Cub...
Leonel Morgado
 
Authoring a personal GPT for your research and practice: How we created the Q...
Authoring a personal GPT for your research and practice: How we created the Q...Authoring a personal GPT for your research and practice: How we created the Q...
Authoring a personal GPT for your research and practice: How we created the Q...
Leonel Morgado
 
Topic: SICKLE CELL DISEASE IN CHILDREN-3.pdf
Topic: SICKLE CELL DISEASE IN CHILDREN-3.pdfTopic: SICKLE CELL DISEASE IN CHILDREN-3.pdf
Topic: SICKLE CELL DISEASE IN CHILDREN-3.pdf
TinyAnderson
 
The binding of cosmological structures by massless topological defects
The binding of cosmological structures by massless topological defectsThe binding of cosmological structures by massless topological defects
The binding of cosmological structures by massless topological defects
Sérgio Sacani
 
如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样
如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样
如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样
yqqaatn0
 

Recently uploaded (20)

Applied Science: Thermodynamics, Laws & Methodology.pdf
Applied Science: Thermodynamics, Laws & Methodology.pdfApplied Science: Thermodynamics, Laws & Methodology.pdf
Applied Science: Thermodynamics, Laws & Methodology.pdf
 
Medical Orthopedic PowerPoint Templates.pptx
Medical Orthopedic PowerPoint Templates.pptxMedical Orthopedic PowerPoint Templates.pptx
Medical Orthopedic PowerPoint Templates.pptx
 
在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样
在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样
在线办理(salfor毕业证书)索尔福德大学毕业证毕业完成信一模一样
 
Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...
Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...
Travis Hills' Endeavors in Minnesota: Fostering Environmental and Economic Pr...
 
Eukaryotic Transcription Presentation.pptx
Eukaryotic Transcription Presentation.pptxEukaryotic Transcription Presentation.pptx
Eukaryotic Transcription Presentation.pptx
 
mô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốt
mô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốtmô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốt
mô tả các thí nghiệm về đánh giá tác động dòng khí hóa sau đốt
 
Deep Software Variability and Frictionless Reproducibility
Deep Software Variability and Frictionless ReproducibilityDeep Software Variability and Frictionless Reproducibility
Deep Software Variability and Frictionless Reproducibility
 
20240520 Planning a Circuit Simulator in JavaScript.pptx
20240520 Planning a Circuit Simulator in JavaScript.pptx20240520 Planning a Circuit Simulator in JavaScript.pptx
20240520 Planning a Circuit Simulator in JavaScript.pptx
 
Compexometric titration/Chelatorphy titration/chelating titration
Compexometric titration/Chelatorphy titration/chelating titrationCompexometric titration/Chelatorphy titration/chelating titration
Compexometric titration/Chelatorphy titration/chelating titration
 
Immersive Learning That Works: Research Grounding and Paths Forward
Immersive Learning That Works: Research Grounding and Paths ForwardImmersive Learning That Works: Research Grounding and Paths Forward
Immersive Learning That Works: Research Grounding and Paths Forward
 
Cytokines and their role in immune regulation.pptx
Cytokines and their role in immune regulation.pptxCytokines and their role in immune regulation.pptx
Cytokines and their role in immune regulation.pptx
 
Randomised Optimisation Algorithms in DAPHNE
Randomised Optimisation Algorithms in DAPHNERandomised Optimisation Algorithms in DAPHNE
Randomised Optimisation Algorithms in DAPHNE
 
Equivariant neural networks and representation theory
Equivariant neural networks and representation theoryEquivariant neural networks and representation theory
Equivariant neural networks and representation theory
 
Phenomics assisted breeding in crop improvement
Phenomics assisted breeding in crop improvementPhenomics assisted breeding in crop improvement
Phenomics assisted breeding in crop improvement
 
aziz sancar nobel prize winner: from mardin to nobel
aziz sancar nobel prize winner: from mardin to nobelaziz sancar nobel prize winner: from mardin to nobel
aziz sancar nobel prize winner: from mardin to nobel
 
Describing and Interpreting an Immersive Learning Case with the Immersion Cub...
Describing and Interpreting an Immersive Learning Case with the Immersion Cub...Describing and Interpreting an Immersive Learning Case with the Immersion Cub...
Describing and Interpreting an Immersive Learning Case with the Immersion Cub...
 
Authoring a personal GPT for your research and practice: How we created the Q...
Authoring a personal GPT for your research and practice: How we created the Q...Authoring a personal GPT for your research and practice: How we created the Q...
Authoring a personal GPT for your research and practice: How we created the Q...
 
Topic: SICKLE CELL DISEASE IN CHILDREN-3.pdf
Topic: SICKLE CELL DISEASE IN CHILDREN-3.pdfTopic: SICKLE CELL DISEASE IN CHILDREN-3.pdf
Topic: SICKLE CELL DISEASE IN CHILDREN-3.pdf
 
The binding of cosmological structures by massless topological defects
The binding of cosmological structures by massless topological defectsThe binding of cosmological structures by massless topological defects
The binding of cosmological structures by massless topological defects
 
如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样
如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样
如何办理(uvic毕业证书)维多利亚大学毕业证本科学位证书原版一模一样
 

NBDT : Neural-backed Decision Tree 2021 ICLR

  • 1. 2021년 5월 23일 딥러닝 논문읽기 모임 이미지 처리팀 : 김병현 안종식 이찬혁 홍은기 NBDT : Neural-backed Decision Tree Wan, Alvin et al. (UC Berkely) ICLR 2021
  • 4. 01. Introduction 1. Necessity of Research [ PCB Inspection System ] Open Defect Image Traditional Image Processing Post-Processing Thresholding Find Condition Open Defect !
  • 5. 01. Introduction 1. Necessity of Research [ PCB Inspection System ] Open Defect Image Deep Learning Based Model Probability of “open” : 99% Probability of “Short” : 1% Probability of “Particle” : 5% Deep Learning is Black Box !
  • 7. 01. Introduction 2. Interpretability 1) Manipulating and Measuring Model Interpretability, Microsoft, 2021 Able to debug Model or Dataset Improving human trust
  • 9. 02. Related Works 1. Saliency Map 1) Grad-CAM : Visual Explanations from Deep Networks via Gradient-based Localization Input Image Saliency map Guided Backpropagation
  • 10. 02. Related Works 1. Saliency Map 1) Grad-CAM : Visual Explanations from Deep Networks via Gradient-based Localization - Output을 도출할 때, Input의 어느 부분이 영향을 얼마나 주는가 ? 검은목논병아리 귀뿔논병아리 Not enough to explain “Why?”
  • 11. 02. Related Works 2. Sequential Decision Processes 1) Decision Tree - Rule Base Model과 같이 의사결정 단계를 작게 쪼게 확인 가능 Oblique Decision Tree Decision Tree
  • 12. Q & A
  • 14. 03. Method 1. Architecture - Classification Head의 마지막 FCN Layer 뒤에 Oblique Decision Tree가 추가적으로 Attach Feature Extractor Classification Head (FCN) Can Apply to Any Classification Models! Induced Hierarchies
  • 15. 03. Method 2. Build Induced Hierarchies 1) Pre-train Model - NBDTs는 Pre-train Model이 필요 : Pretrain Model의 마지막 FCN Layer Weight를 이용해 Hierarchies 구성 Build Hierarchies Inference
  • 16. 03. Method 2. Build Induced Hierarchies 2) Agglomerative Clustering - build Hierarchies
  • 17. 03. Method 2. Build Induced Hierarchies 2) Agglomerative Clustering - build Hierarchies k Weight Space Clustering Repeat
  • 18. 03. Method 2. Build Induced Hierarchies 2) Agglomerative Clustering - build Hierarchies Problem 1. Problem 2. ? ?
  • 19. 03. Method 2. Build Induced Hierarchies 3) WordNet : 단어들을 계층적으로 구조화 해 놓은 것 - Clustering 된 두 Class들의 공통 조상을 WordNet에서 찾음 [ WordNet ] CIFAR-100
  • 20. 03. Method 2. Build Induced Hierarchies 4) Set Weight of Induced Hierarchies Step A,B) Load Pretrain Model’s Weight & Allocate Weight to each Leaf Nodes Step C) Set Parent Vector : Average of Child Node Step D) Repeat Until Set all node’s weight 𝑣𝑒𝑐𝑡𝑜𝑟 𝑛1 d k Average
  • 21. 03. Method 3. Inference 1) Soft Decision Tree vs Hard Decision Tree Inner Product - Probability of class K - Final Prediction “분기” 개념보다 “Path” 개념에 가깝다! Softmax
  • 22. 03. Method 4. Loss 1) Tree Supervision Loss
  • 23. 03. Method 4. Loss 1) Tree Supervision Loss Feature Extractor Classification Head (FCN) Induced Hierarchies 0  0.5 1  0
  • 24. Q & A
  • 26. 04. Experiments 1. Accuracy 1) Small-scale Datasets
  • 27. 04. Experiments 1. Accuracy 2) Large-scale Datasets
  • 28. 04. Experiments 2. Ablation Study 1) Hierarchies
  • 29. 04. Experiments 2. Ablation Study 2) Losses
  • 30. 04. Experiments 3. Interpretability : Survey 1) 방법 : Saliency Map과 NBDT 설명이 주어졌을 때, 어느 것이 더 옳게 찾는지 Survey 2) 종류 a) 주어진 설명을 통해서 Model의 오탐지를 찾을 수 있는가? - 결과 : 600개의 응답 중 Saliency Map은 87개 정답 / NBDT는 237개 정답 b) 어떤 모델의 판단을 더 신뢰하는가? - 결과 : 374개의 응답 중 65.9%가 NBDT를 더 신뢰함
  • 31. 04. Experiments 3. Interpretability : Survey c) 판단이 어려운 상황에서 모델의 예측을 신뢰할 수 있는가? - 매우 Blur한 이미지와 각 모델이 예측한 결과를 제시 > 이때, Prediction의 결과는 다름 (NBDT가 정답 : Saliency Map가 정답 : 둘 다 틀림 = 3 : 3 : 4) - 결과 : 600개의 응답 중 167개 Saliency Map의 결과에 동의, 312개 NBDT의 결과에 동의, 나머지는 둘다 반대 > 600개의 결과 중 255개의 응답이 정답을 맞춤
  • 32. 04. Experiments 3. Interpretability : Analysis 2) Model이 헷갈리는 이미지를 찾는데 사용될 수 있는 가? - 헷갈리는 문제에서, 확실한 부분은 높은 Probability를, 헷갈리는 부분에서는 낮은 Probability를 보인다 - 위를 이용하여 “Path Entropy”를 계산하여 헷갈리는 Image를 구분해낼 수 있다.
  • 33. 04. Experiments 3. Interpretability : Analysis 3) Model이 헷갈리는 이미지를 찾는데 사용될 수 있는 가? - ImageNet 내의 Multi Target Image를 분류하는데 사용 가능
  • 35. 05. Conclusion 1. Contribution 1) Accuracy를 유지하며 Interpretability를 향상 2) Grad-CAM 기법에 비해 Interpretability가 높음 3) 모든 Model에 적용 가능 2. Discussion 1) Good - Survey 기반으로 Human Trust를 수치화 한 것은 인상적 - ImageNet의 잘못된 Image를 분리해 낼 수 있다 : Sub-Result 2) Bad - 논문 전반적으로 혼동이 올 수 있는 언어 사용 / 설명 Skip이 많음
  • 36. Q & A