SlideShare a Scribd company logo
1 of 21
Download to read offline
2018. 10. 24.
의료영상 분석 및 처리
- Auto ML -
이 동 헌
https://www.slideshare.net/KihoSuh/neural-architecture-search-with-reinforcement-learning-76883153
v 큰 구조: fixed
• Normal / Reduction Cell #
• B = 5
https://www.slideshare.net/KihoSuh/neural-architecture-search-with-reinforcement-learning-76883153
v세부 사항: Searching (Normal Cell / Reduction Cell 내부)
The 13 available operations
Softmax Skip Connection
• Penn Treebank
The CIFAR-10 dataset consists of 60000 32x32 colour
images in 10 classes, with 6000 images per class. There
are 50000 training images and 10000 test images.
• CIFAR-10
Penn Treebank dataset, known as PTB dataset, is widely
used in machine learning of NLP (Natural Language
Processing) research
v Datasets
Keras
Tensorflow
NAS is computationally expensive and time consuming, e.g. Zoph et al. (2018) use 450 GPUs for 3-4 days
(i.e. 32,400-43,200 GPU hours)
• We observe that the computational bottleneck of NAS is the training of each child model to
convergence, only to measure its accuracy whilst throwing away all the trained weights.
• The main contribution of this work is to improve the efficiency of NAS by forcing all child models
to share weights to eschew training each child model from scratch to convergence.
Importantly, in all of our experiments, for which we use a single Nvidia GTX 1080Ti GPU, the search for
architectures takes less than 16 hours. Compared to NAS, this is a reduction of GPU-hours by more than 1000x.
uDirected Acyclic Graph (DAG)
§ Node
• Local Computation
• Own Parameters (activate시, 사용)
§ Edge
• Flow of information
• Determined by a controller (red)
Input
Output
→ Search space 내의 모든 child model의 parameter sharing
→ RNN cell의 1node 배치와 2operation을 같이 학습 (유연함)
(↔ NAS: 사용자가 node의 배치 등을 정해주고, 각 노드의 operation만 학습)
https://jayhey.github.io/deep%20learning/2018/03/15/ENAS/
(1) Macro Search : 전체 구조 Search (7시간)
The 6 available operations (< NAS)
• Convolution with kernel size 3 × 3 and 5 × 5.
• Depthwise-Separable Convolution with kernel size 3 × 3 and 5 × 5.
• Average Pooling / Max pooling with kernel size 3 × 3.
6L×2L(L-1)/2 개
(L=12, 1.6×1029 개 후보)
(2) Micro Search
: Cell 단위로 Search하고나서 합침 (11.5시간)
The 5 available operations (< NAS)
• Identity
• Separable Convolution with kernel size 3 × 3 and 5 × 5.
• Average Pooling / Max pooling with kernel size 3 × 3.
(5×(B-2)!)4 개
(B=7, 1.3×1011 개 후보)
https://github.com/melodyguan/enas
Normal Cell
Reduction Cell
ModuLab DLC-Medical3
ModuLab DLC-Medical3
ModuLab DLC-Medical3

More Related Content

What's hot

increasing the action gap - new operators for reinforcement learning
increasing the action gap - new operators for reinforcement learningincreasing the action gap - new operators for reinforcement learning
increasing the action gap - new operators for reinforcement learningRyo Iwaki
 
Back propagation
Back propagationBack propagation
Back propagationSan Kim
 
Getting your hands dirty with deep learning in java
Getting your hands dirty with deep learning in javaGetting your hands dirty with deep learning in java
Getting your hands dirty with deep learning in javaDave Snowdon
 
딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)
딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)
딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)Hansol Kang
 
Complex models in ecology: challenges and solutions
Complex models in ecology: challenges and solutionsComplex models in ecology: challenges and solutions
Complex models in ecology: challenges and solutionsPeter Solymos
 
Photo-realistic Single Image Super-resolution using a Generative Adversarial ...
Photo-realistic Single Image Super-resolution using a Generative Adversarial ...Photo-realistic Single Image Super-resolution using a Generative Adversarial ...
Photo-realistic Single Image Super-resolution using a Generative Adversarial ...Hansol Kang
 
[251] implementing deep learning using cu dnn
[251] implementing deep learning using cu dnn[251] implementing deep learning using cu dnn
[251] implementing deep learning using cu dnnNAVER D2
 
Deep learning study 2
Deep learning study 2Deep learning study 2
Deep learning study 2San Kim
 
TensorFlow 深度學習快速上手班--電腦視覺應用
TensorFlow 深度學習快速上手班--電腦視覺應用TensorFlow 深度學習快速上手班--電腦視覺應用
TensorFlow 深度學習快速上手班--電腦視覺應用Mark Chang
 
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...Deep Learning JP
 
Intro to TensorFlow and PyTorch Workshop at Tubular Labs
Intro to TensorFlow and PyTorch Workshop at Tubular LabsIntro to TensorFlow and PyTorch Workshop at Tubular Labs
Intro to TensorFlow and PyTorch Workshop at Tubular LabsKendall
 
H2O Deep Learning at Next.ML
H2O Deep Learning at Next.MLH2O Deep Learning at Next.ML
H2O Deep Learning at Next.MLSri Ambati
 
Gossip-based resource allocation for green computing in large clouds
Gossip-based resource allocation for green computing in large cloudsGossip-based resource allocation for green computing in large clouds
Gossip-based resource allocation for green computing in large cloudsRerngvit Yanggratoke
 
Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016
Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016
Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016MLconf
 
An introduction to Deep Learning with Apache MXNet (November 2017)
An introduction to Deep Learning with Apache MXNet (November 2017)An introduction to Deep Learning with Apache MXNet (November 2017)
An introduction to Deep Learning with Apache MXNet (November 2017)Julien SIMON
 
How to apply deep learning to 3 d objects
How to apply deep learning to 3 d objectsHow to apply deep learning to 3 d objects
How to apply deep learning to 3 d objectsOgushi Masaya
 
Deep Learning for Developers
Deep Learning for DevelopersDeep Learning for Developers
Deep Learning for DevelopersJulien SIMON
 

What's hot (20)

increasing the action gap - new operators for reinforcement learning
increasing the action gap - new operators for reinforcement learningincreasing the action gap - new operators for reinforcement learning
increasing the action gap - new operators for reinforcement learning
 
Slide tesi
Slide tesiSlide tesi
Slide tesi
 
Back propagation
Back propagationBack propagation
Back propagation
 
Getting your hands dirty with deep learning in java
Getting your hands dirty with deep learning in javaGetting your hands dirty with deep learning in java
Getting your hands dirty with deep learning in java
 
딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)
딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)
딥러닝 중급 - AlexNet과 VggNet (Basic of DCNN : AlexNet and VggNet)
 
Complex models in ecology: challenges and solutions
Complex models in ecology: challenges and solutionsComplex models in ecology: challenges and solutions
Complex models in ecology: challenges and solutions
 
Photo-realistic Single Image Super-resolution using a Generative Adversarial ...
Photo-realistic Single Image Super-resolution using a Generative Adversarial ...Photo-realistic Single Image Super-resolution using a Generative Adversarial ...
Photo-realistic Single Image Super-resolution using a Generative Adversarial ...
 
GAN Evaluation
GAN EvaluationGAN Evaluation
GAN Evaluation
 
[251] implementing deep learning using cu dnn
[251] implementing deep learning using cu dnn[251] implementing deep learning using cu dnn
[251] implementing deep learning using cu dnn
 
Deep learning study 2
Deep learning study 2Deep learning study 2
Deep learning study 2
 
TensorFlow 深度學習快速上手班--電腦視覺應用
TensorFlow 深度學習快速上手班--電腦視覺應用TensorFlow 深度學習快速上手班--電腦視覺應用
TensorFlow 深度學習快速上手班--電腦視覺應用
 
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
【DL輪読会】Toward Fast and Stabilized GAN Training for Highfidelity Few-shot Imag...
 
Intro to TensorFlow and PyTorch Workshop at Tubular Labs
Intro to TensorFlow and PyTorch Workshop at Tubular LabsIntro to TensorFlow and PyTorch Workshop at Tubular Labs
Intro to TensorFlow and PyTorch Workshop at Tubular Labs
 
H2O Deep Learning at Next.ML
H2O Deep Learning at Next.MLH2O Deep Learning at Next.ML
H2O Deep Learning at Next.ML
 
Gossip-based resource allocation for green computing in large clouds
Gossip-based resource allocation for green computing in large cloudsGossip-based resource allocation for green computing in large clouds
Gossip-based resource allocation for green computing in large clouds
 
Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016
Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016
Alex Smola, Director of Machine Learning, AWS/Amazon, at MLconf SF 2016
 
An introduction to Deep Learning with Apache MXNet (November 2017)
An introduction to Deep Learning with Apache MXNet (November 2017)An introduction to Deep Learning with Apache MXNet (November 2017)
An introduction to Deep Learning with Apache MXNet (November 2017)
 
Profiling in Python
Profiling in PythonProfiling in Python
Profiling in Python
 
How to apply deep learning to 3 d objects
How to apply deep learning to 3 d objectsHow to apply deep learning to 3 d objects
How to apply deep learning to 3 d objects
 
Deep Learning for Developers
Deep Learning for DevelopersDeep Learning for Developers
Deep Learning for Developers
 

Similar to ModuLab DLC-Medical3

Distributed DNN training: Infrastructure, challenges, and lessons learned
Distributed DNN training: Infrastructure, challenges, and lessons learnedDistributed DNN training: Infrastructure, challenges, and lessons learned
Distributed DNN training: Infrastructure, challenges, and lessons learnedWee Hyong Tok
 
Using Elasticsearch as the Primary Data Store
Using Elasticsearch as the Primary Data StoreUsing Elasticsearch as the Primary Data Store
Using Elasticsearch as the Primary Data StoreVolkan Yazıcı
 
XSEDE April 2017
XSEDE April 2017XSEDE April 2017
XSEDE April 2017SciCompIIT
 
Efficient Neural Architecture Search via Parameter Sharing
Efficient Neural Architecture Search via Parameter SharingEfficient Neural Architecture Search via Parameter Sharing
Efficient Neural Architecture Search via Parameter SharingJinwon Lee
 
Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)
Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)
Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)Kohei KaiGai
 
Machine learning at Scale with Apache Spark
Machine learning at Scale with Apache SparkMachine learning at Scale with Apache Spark
Machine learning at Scale with Apache SparkMartin Zapletal
 
SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...
SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...
SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...Chester Chen
 
Terabyte-scale image similarity search: experience and best practice
Terabyte-scale image similarity search: experience and best practiceTerabyte-scale image similarity search: experience and best practice
Terabyte-scale image similarity search: experience and best practiceDenis Shestakov
 
In-memory Data Management Trends & Techniques
In-memory Data Management Trends & TechniquesIn-memory Data Management Trends & Techniques
In-memory Data Management Trends & TechniquesHazelcast
 
Resource-Efficient Deep Learning Model Selection on Apache Spark
Resource-Efficient Deep Learning Model Selection on Apache SparkResource-Efficient Deep Learning Model Selection on Apache Spark
Resource-Efficient Deep Learning Model Selection on Apache SparkDatabricks
 
Java Thread and Process Performance for Parallel Machine Learning on Multicor...
Java Thread and Process Performance for Parallel Machine Learning on Multicor...Java Thread and Process Performance for Parallel Machine Learning on Multicor...
Java Thread and Process Performance for Parallel Machine Learning on Multicor...Saliya Ekanayake
 
Scalable high-dimensional indexing with Hadoop
Scalable high-dimensional indexing with HadoopScalable high-dimensional indexing with Hadoop
Scalable high-dimensional indexing with HadoopDenis Shestakov
 
Fast and Scalable Python
Fast and Scalable PythonFast and Scalable Python
Fast and Scalable PythonTravis Oliphant
 
KIISE:SIGDB Workshop presentation.
KIISE:SIGDB Workshop presentation.KIISE:SIGDB Workshop presentation.
KIISE:SIGDB Workshop presentation.Kyong-Ha Lee
 
Yarn 3.1 and Beyond in ApacheCon North America 2018
Yarn 3.1 and Beyond in ApacheCon North America 2018Yarn 3.1 and Beyond in ApacheCon North America 2018
Yarn 3.1 and Beyond in ApacheCon North America 2018Naganarasimha Garla
 
Benchmarking Solr Performance at Scale
Benchmarking Solr Performance at ScaleBenchmarking Solr Performance at Scale
Benchmarking Solr Performance at Scalethelabdude
 

Similar to ModuLab DLC-Medical3 (20)

Distributed DNN training: Infrastructure, challenges, and lessons learned
Distributed DNN training: Infrastructure, challenges, and lessons learnedDistributed DNN training: Infrastructure, challenges, and lessons learned
Distributed DNN training: Infrastructure, challenges, and lessons learned
 
Using Elasticsearch as the Primary Data Store
Using Elasticsearch as the Primary Data StoreUsing Elasticsearch as the Primary Data Store
Using Elasticsearch as the Primary Data Store
 
XSEDE April 2017
XSEDE April 2017XSEDE April 2017
XSEDE April 2017
 
GIST AI-X Computing Cluster
GIST AI-X Computing ClusterGIST AI-X Computing Cluster
GIST AI-X Computing Cluster
 
Efficient Neural Architecture Search via Parameter Sharing
Efficient Neural Architecture Search via Parameter SharingEfficient Neural Architecture Search via Parameter Sharing
Efficient Neural Architecture Search via Parameter Sharing
 
Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)
Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)
Technology Updates of PG-Strom at Aug-2014 (PGUnconf@Tokyo)
 
Machine learning at Scale with Apache Spark
Machine learning at Scale with Apache SparkMachine learning at Scale with Apache Spark
Machine learning at Scale with Apache Spark
 
SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...
SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...
SF Big Analytics & SF Machine Learning Meetup: Machine Learning at the Limit ...
 
Terabyte-scale image similarity search: experience and best practice
Terabyte-scale image similarity search: experience and best practiceTerabyte-scale image similarity search: experience and best practice
Terabyte-scale image similarity search: experience and best practice
 
In-memory Data Management Trends & Techniques
In-memory Data Management Trends & TechniquesIn-memory Data Management Trends & Techniques
In-memory Data Management Trends & Techniques
 
Manycores for the Masses
Manycores for the MassesManycores for the Masses
Manycores for the Masses
 
Resource-Efficient Deep Learning Model Selection on Apache Spark
Resource-Efficient Deep Learning Model Selection on Apache SparkResource-Efficient Deep Learning Model Selection on Apache Spark
Resource-Efficient Deep Learning Model Selection on Apache Spark
 
Java Thread and Process Performance for Parallel Machine Learning on Multicor...
Java Thread and Process Performance for Parallel Machine Learning on Multicor...Java Thread and Process Performance for Parallel Machine Learning on Multicor...
Java Thread and Process Performance for Parallel Machine Learning on Multicor...
 
Scalable high-dimensional indexing with Hadoop
Scalable high-dimensional indexing with HadoopScalable high-dimensional indexing with Hadoop
Scalable high-dimensional indexing with Hadoop
 
Lrz kurs: big data analysis
Lrz kurs: big data analysisLrz kurs: big data analysis
Lrz kurs: big data analysis
 
20080528dublinpt3
20080528dublinpt320080528dublinpt3
20080528dublinpt3
 
Fast and Scalable Python
Fast and Scalable PythonFast and Scalable Python
Fast and Scalable Python
 
KIISE:SIGDB Workshop presentation.
KIISE:SIGDB Workshop presentation.KIISE:SIGDB Workshop presentation.
KIISE:SIGDB Workshop presentation.
 
Yarn 3.1 and Beyond in ApacheCon North America 2018
Yarn 3.1 and Beyond in ApacheCon North America 2018Yarn 3.1 and Beyond in ApacheCon North America 2018
Yarn 3.1 and Beyond in ApacheCon North America 2018
 
Benchmarking Solr Performance at Scale
Benchmarking Solr Performance at ScaleBenchmarking Solr Performance at Scale
Benchmarking Solr Performance at Scale
 

More from Dongheon Lee

Workshop 210417 dhlee
Workshop 210417 dhleeWorkshop 210417 dhlee
Workshop 210417 dhleeDongheon Lee
 
Pixel RNN to Pixel CNN++
Pixel RNN to Pixel CNN++Pixel RNN to Pixel CNN++
Pixel RNN to Pixel CNN++Dongheon Lee
 
ModuLab DLC-Medical5
ModuLab DLC-Medical5ModuLab DLC-Medical5
ModuLab DLC-Medical5Dongheon Lee
 
ModuLab DLC-Medical4
ModuLab DLC-Medical4ModuLab DLC-Medical4
ModuLab DLC-Medical4Dongheon Lee
 
ModuLab DLC-Medical1
ModuLab DLC-Medical1ModuLab DLC-Medical1
ModuLab DLC-Medical1Dongheon Lee
 
Deep Learning for AI (1)
Deep Learning for AI (1)Deep Learning for AI (1)
Deep Learning for AI (1)Dongheon Lee
 

More from Dongheon Lee (7)

Workshop 210417 dhlee
Workshop 210417 dhleeWorkshop 210417 dhlee
Workshop 210417 dhlee
 
BeautyGlow
BeautyGlowBeautyGlow
BeautyGlow
 
Pixel RNN to Pixel CNN++
Pixel RNN to Pixel CNN++Pixel RNN to Pixel CNN++
Pixel RNN to Pixel CNN++
 
ModuLab DLC-Medical5
ModuLab DLC-Medical5ModuLab DLC-Medical5
ModuLab DLC-Medical5
 
ModuLab DLC-Medical4
ModuLab DLC-Medical4ModuLab DLC-Medical4
ModuLab DLC-Medical4
 
ModuLab DLC-Medical1
ModuLab DLC-Medical1ModuLab DLC-Medical1
ModuLab DLC-Medical1
 
Deep Learning for AI (1)
Deep Learning for AI (1)Deep Learning for AI (1)
Deep Learning for AI (1)
 

Recently uploaded

SHOCK (Medical SURGICAL BASED EDITION)).pptx
SHOCK (Medical SURGICAL BASED EDITION)).pptxSHOCK (Medical SURGICAL BASED EDITION)).pptx
SHOCK (Medical SURGICAL BASED EDITION)).pptxAbhishek943418
 
Biomechanics- Shoulder Joint!!!!!!!!!!!!
Biomechanics- Shoulder Joint!!!!!!!!!!!!Biomechanics- Shoulder Joint!!!!!!!!!!!!
Biomechanics- Shoulder Joint!!!!!!!!!!!!ibtesaam huma
 
Primary headache and facial pain. (2024)
Primary headache and facial pain. (2024)Primary headache and facial pain. (2024)
Primary headache and facial pain. (2024)Mohamed Rizk Khodair
 
Basic principles involved in the traditional systems of medicine PDF.pdf
Basic principles involved in the traditional systems of medicine PDF.pdfBasic principles involved in the traditional systems of medicine PDF.pdf
Basic principles involved in the traditional systems of medicine PDF.pdfDivya Kanojiya
 
Presentation on Parasympathetic Nervous System
Presentation on Parasympathetic Nervous SystemPresentation on Parasympathetic Nervous System
Presentation on Parasympathetic Nervous SystemPrerana Jadhav
 
COVID-19 (NOVEL CORONA VIRUS DISEASE PANDEMIC ).pptx
COVID-19  (NOVEL CORONA  VIRUS DISEASE PANDEMIC ).pptxCOVID-19  (NOVEL CORONA  VIRUS DISEASE PANDEMIC ).pptx
COVID-19 (NOVEL CORONA VIRUS DISEASE PANDEMIC ).pptxBibekananda shah
 
Valproic Acid. (VPA). Antiseizure medication
Valproic Acid.  (VPA). Antiseizure medicationValproic Acid.  (VPA). Antiseizure medication
Valproic Acid. (VPA). Antiseizure medicationMohamadAlhes
 
Phytochemical Investigation of Drugs PDF.pdf
Phytochemical Investigation of Drugs PDF.pdfPhytochemical Investigation of Drugs PDF.pdf
Phytochemical Investigation of Drugs PDF.pdfDivya Kanojiya
 
SYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptx
SYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptxSYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptx
SYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptxdrashraf369
 
Big Data Analysis Suggests COVID Vaccination Increases Excess Mortality Of ...
Big Data Analysis Suggests COVID  Vaccination Increases Excess Mortality Of  ...Big Data Analysis Suggests COVID  Vaccination Increases Excess Mortality Of  ...
Big Data Analysis Suggests COVID Vaccination Increases Excess Mortality Of ...sdateam0
 
Plant Fibres used as Surgical Dressings PDF.pdf
Plant Fibres used as Surgical Dressings PDF.pdfPlant Fibres used as Surgical Dressings PDF.pdf
Plant Fibres used as Surgical Dressings PDF.pdfDivya Kanojiya
 
Chronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptx
Chronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptxChronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptx
Chronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptxSasikiranMarri
 
epilepsy and status epilepticus for undergraduate.pptx
epilepsy and status epilepticus  for undergraduate.pptxepilepsy and status epilepticus  for undergraduate.pptx
epilepsy and status epilepticus for undergraduate.pptxMohamed Rizk Khodair
 
Presentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptx
Presentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptxPresentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptx
Presentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptxpdamico1
 
ANEMIA IN PREGNANCY by Dr. Akebom Kidanemariam
ANEMIA IN PREGNANCY by Dr. Akebom KidanemariamANEMIA IN PREGNANCY by Dr. Akebom Kidanemariam
ANEMIA IN PREGNANCY by Dr. Akebom KidanemariamAkebom Gebremichael
 
SGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdf
SGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdfSGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdf
SGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdfHongBiThi1
 
SCHOOL HEALTH SERVICES.pptx made by Sapna Thakur
SCHOOL HEALTH SERVICES.pptx made by Sapna ThakurSCHOOL HEALTH SERVICES.pptx made by Sapna Thakur
SCHOOL HEALTH SERVICES.pptx made by Sapna ThakurSapna Thakur
 
Hypersensitivity and its classification .pptx
Hypersensitivity and its classification .pptxHypersensitivity and its classification .pptx
Hypersensitivity and its classification .pptxAkshay Shetty
 
Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...
Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...
Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...MehranMouzam
 

Recently uploaded (20)

SHOCK (Medical SURGICAL BASED EDITION)).pptx
SHOCK (Medical SURGICAL BASED EDITION)).pptxSHOCK (Medical SURGICAL BASED EDITION)).pptx
SHOCK (Medical SURGICAL BASED EDITION)).pptx
 
Biomechanics- Shoulder Joint!!!!!!!!!!!!
Biomechanics- Shoulder Joint!!!!!!!!!!!!Biomechanics- Shoulder Joint!!!!!!!!!!!!
Biomechanics- Shoulder Joint!!!!!!!!!!!!
 
Primary headache and facial pain. (2024)
Primary headache and facial pain. (2024)Primary headache and facial pain. (2024)
Primary headache and facial pain. (2024)
 
Basic principles involved in the traditional systems of medicine PDF.pdf
Basic principles involved in the traditional systems of medicine PDF.pdfBasic principles involved in the traditional systems of medicine PDF.pdf
Basic principles involved in the traditional systems of medicine PDF.pdf
 
Presentation on Parasympathetic Nervous System
Presentation on Parasympathetic Nervous SystemPresentation on Parasympathetic Nervous System
Presentation on Parasympathetic Nervous System
 
COVID-19 (NOVEL CORONA VIRUS DISEASE PANDEMIC ).pptx
COVID-19  (NOVEL CORONA  VIRUS DISEASE PANDEMIC ).pptxCOVID-19  (NOVEL CORONA  VIRUS DISEASE PANDEMIC ).pptx
COVID-19 (NOVEL CORONA VIRUS DISEASE PANDEMIC ).pptx
 
Valproic Acid. (VPA). Antiseizure medication
Valproic Acid.  (VPA). Antiseizure medicationValproic Acid.  (VPA). Antiseizure medication
Valproic Acid. (VPA). Antiseizure medication
 
Phytochemical Investigation of Drugs PDF.pdf
Phytochemical Investigation of Drugs PDF.pdfPhytochemical Investigation of Drugs PDF.pdf
Phytochemical Investigation of Drugs PDF.pdf
 
JANGAMA VISHA .pptx-
JANGAMA VISHA .pptx-JANGAMA VISHA .pptx-
JANGAMA VISHA .pptx-
 
SYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptx
SYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptxSYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptx
SYNDESMOTIC INJURY- ANATOMICAL REPAIR.pptx
 
Big Data Analysis Suggests COVID Vaccination Increases Excess Mortality Of ...
Big Data Analysis Suggests COVID  Vaccination Increases Excess Mortality Of  ...Big Data Analysis Suggests COVID  Vaccination Increases Excess Mortality Of  ...
Big Data Analysis Suggests COVID Vaccination Increases Excess Mortality Of ...
 
Plant Fibres used as Surgical Dressings PDF.pdf
Plant Fibres used as Surgical Dressings PDF.pdfPlant Fibres used as Surgical Dressings PDF.pdf
Plant Fibres used as Surgical Dressings PDF.pdf
 
Chronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptx
Chronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptxChronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptx
Chronic-Fatigue-Syndrome-CFS-Understanding-a-Complex-Disorder.pptx
 
epilepsy and status epilepticus for undergraduate.pptx
epilepsy and status epilepticus  for undergraduate.pptxepilepsy and status epilepticus  for undergraduate.pptx
epilepsy and status epilepticus for undergraduate.pptx
 
Presentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptx
Presentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptxPresentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptx
Presentation for Bella Mahl 2024-03-28-24-MW-Overview-Bella.pptx
 
ANEMIA IN PREGNANCY by Dr. Akebom Kidanemariam
ANEMIA IN PREGNANCY by Dr. Akebom KidanemariamANEMIA IN PREGNANCY by Dr. Akebom Kidanemariam
ANEMIA IN PREGNANCY by Dr. Akebom Kidanemariam
 
SGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdf
SGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdfSGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdf
SGK HÓA SINH NĂNG LƯỢNG SINH HỌC 2006.pdf
 
SCHOOL HEALTH SERVICES.pptx made by Sapna Thakur
SCHOOL HEALTH SERVICES.pptx made by Sapna ThakurSCHOOL HEALTH SERVICES.pptx made by Sapna Thakur
SCHOOL HEALTH SERVICES.pptx made by Sapna Thakur
 
Hypersensitivity and its classification .pptx
Hypersensitivity and its classification .pptxHypersensitivity and its classification .pptx
Hypersensitivity and its classification .pptx
 
Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...
Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...
Study on the Impact of FOCUS-PDCA Management Model on the Disinfection Qualit...
 

ModuLab DLC-Medical3

  • 1. 2018. 10. 24. 의료영상 분석 및 처리 - Auto ML - 이 동 헌
  • 2.
  • 3.
  • 5. v 큰 구조: fixed • Normal / Reduction Cell # • B = 5
  • 6.
  • 7. https://www.slideshare.net/KihoSuh/neural-architecture-search-with-reinforcement-learning-76883153 v세부 사항: Searching (Normal Cell / Reduction Cell 내부) The 13 available operations Softmax Skip Connection
  • 8. • Penn Treebank The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes, with 6000 images per class. There are 50000 training images and 10000 test images. • CIFAR-10 Penn Treebank dataset, known as PTB dataset, is widely used in machine learning of NLP (Natural Language Processing) research v Datasets
  • 10. NAS is computationally expensive and time consuming, e.g. Zoph et al. (2018) use 450 GPUs for 3-4 days (i.e. 32,400-43,200 GPU hours) • We observe that the computational bottleneck of NAS is the training of each child model to convergence, only to measure its accuracy whilst throwing away all the trained weights. • The main contribution of this work is to improve the efficiency of NAS by forcing all child models to share weights to eschew training each child model from scratch to convergence. Importantly, in all of our experiments, for which we use a single Nvidia GTX 1080Ti GPU, the search for architectures takes less than 16 hours. Compared to NAS, this is a reduction of GPU-hours by more than 1000x.
  • 11. uDirected Acyclic Graph (DAG) § Node • Local Computation • Own Parameters (activate시, 사용) § Edge • Flow of information • Determined by a controller (red) Input Output → Search space 내의 모든 child model의 parameter sharing → RNN cell의 1node 배치와 2operation을 같이 학습 (유연함) (↔ NAS: 사용자가 node의 배치 등을 정해주고, 각 노드의 operation만 학습) https://jayhey.github.io/deep%20learning/2018/03/15/ENAS/
  • 12. (1) Macro Search : 전체 구조 Search (7시간) The 6 available operations (< NAS) • Convolution with kernel size 3 × 3 and 5 × 5. • Depthwise-Separable Convolution with kernel size 3 × 3 and 5 × 5. • Average Pooling / Max pooling with kernel size 3 × 3. 6L×2L(L-1)/2 개 (L=12, 1.6×1029 개 후보)
  • 13.
  • 14. (2) Micro Search : Cell 단위로 Search하고나서 합침 (11.5시간) The 5 available operations (< NAS) • Identity • Separable Convolution with kernel size 3 × 3 and 5 × 5. • Average Pooling / Max pooling with kernel size 3 × 3. (5×(B-2)!)4 개 (B=7, 1.3×1011 개 후보)
  • 15.
  • 16.