SlideShare une entreprise Scribd logo
1  sur  1
Passionate
on Parallel
REU 2015
This REU is co-funded by the ASSURE program of the Department of
Defense in partnership with the National Science Foundation REU Site
Program under Award No. 1263145.
Accelerating Convolution Neural Network Learning with MPI
Dustyn Tubbs, Saginaw Valley State University Patrick Streifel, St. Mary’s College of Maryland Advisors: Dr. Deming Chen and Ashutosh Dhar
Convolution Neural Networks (CNNs) are a type of feed-
forward neural network specialized for classifying objects
in images. CNNs guess the class of an object in an image by
applying multiple convolution filters across regions of the
image. To train a CNN, these filters are adjusted based on a
guess’s error, so that the updated filters produces slightly
more accurate classifications. This method of supervised
learning typically takes place over tens of thousands of
labeled images.
Background
Having finished data-parallelism in programming,
our future work is based on acquiring training-
time metrics for published architectures (such as
the GoogLeNet architecture in [1]) and
implementing model parallelism as another form
of CNN parallelism. After this is accomplished, we
intend to release a package to the Torch7
community which will let them easily perform
both model and data parallelism in their research.
Future Works
Ever since 2012, when the first CNN to win the biannual
ImageNet Challenge (a visual recognition contest)
outperformed its closest competitor by 11%, CNNs have
become the state of the art for object identification. They
are being used today by big names like Google and
Facebook to improve image searches, auto-tag people in
photograph, and countless other applications.
Given the rapid development in CNN research, the need to
exploit parallelization techniques to allow for both larger
networks and faster training is apparent.
Motivation
[1] Szegedy, Christian, et al. "Going deeper with
convolutions." arXiv preprint arXiv:1409.4842 (2014).
[2] Krizhevsky, Alex. "One weird trick for parallelizing
convolutional neural networks." arXiv preprint
arXiv:1404.5997 (2014).
This research project could not have been accomplished without
the dedicated support of our mentors Deming Chen and
Ashutosh Dhar, our REU leaders and organizers Jill Peckham,
Craig Ziles, and Mathew West, and the University of Illinois at
Urbana-Champaign. Special thanks to the Circuits Research
Group at CSL for hosting us for the duration of our project.
References and
Acknowledgements
Data-Parallelism in Convolution Neural Networks is accomplished by performing
distribution of both the network and the data across multiple nodes. Traditionally, a
CNN is trained serially on a single node, which loads the entire data set (as seen to
the right). Compare this structure with the below structure, which distributes both
the data and the network. Instead of training on one image at a time, the CNN can
train on each node-local chunk of training data and communicate the necessary
changes to every network copy across the network. This is superior in that it lets the
researcher quickly train the network, adjust the topology, and train again.
Data-ParallelismWe used Torch7, a framework for working with CNNs.
During our research, we discovered that Torch7 had no
such convenient utilities for MPI parallelization. So we
decided that our goal should be to implement MPI
parallelization for Torch CNNs.
Goal
There are two paradigms for CNN parallelization: model and data parallelism. In
model parallelism, the CNN architecture is divided among several workers and
trained on the same batch of images. This technique allows you to train on much
larger CNNs. In data parallelism, identical copies of the CNN are maintained by each
worker and trained in parallel on separate batches of images. This allows you to train
a CNN on many more images at one time.
Parallelization Methods

Contenu connexe

Tendances

Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...
Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...
Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...multimediaeval
 
Machine Learning in Healthcare Diagnostics
Machine Learning in Healthcare DiagnosticsMachine Learning in Healthcare Diagnostics
Machine Learning in Healthcare DiagnosticsLarry Smarr
 
Building the Pacific Research Platform: Supernetworks for Big Data Science
Building the Pacific Research Platform: Supernetworks for Big Data ScienceBuilding the Pacific Research Platform: Supernetworks for Big Data Science
Building the Pacific Research Platform: Supernetworks for Big Data ScienceLarry Smarr
 
"Methods for Understanding How Deep Neural Networks Work," a Presentation fro...
"Methods for Understanding How Deep Neural Networks Work," a Presentation fro..."Methods for Understanding How Deep Neural Networks Work," a Presentation fro...
"Methods for Understanding How Deep Neural Networks Work," a Presentation fro...Edge AI and Vision Alliance
 
Mind mapping and Its Applications, Introduction to Context Trees
Mind mapping and Its Applications, Introduction to Context TreesMind mapping and Its Applications, Introduction to Context Trees
Mind mapping and Its Applications, Introduction to Context TreesSunayana Gawde
 
Metron seas collaboration
Metron seas collaborationMetron seas collaboration
Metron seas collaborationikekala
 
A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...
A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...
A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...Databricks
 
Quantum computing and machine learning overview
Quantum computing and machine learning overviewQuantum computing and machine learning overview
Quantum computing and machine learning overviewColleen Farrelly
 
A New Approach to Video Compression using 3D-DCT
A New Approach to Video Compression using 3D-DCTA New Approach to Video Compression using 3D-DCT
A New Approach to Video Compression using 3D-DCTShanmuga Priyan Thiagarajan
 
Abstract imakawa
Abstract imakawaAbstract imakawa
Abstract imakawaharmonylab
 

Tendances (13)

Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...
Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...
Classification of Strokes in Table Tennis with a Three Stream Spatio-Temporal...
 
September 28, Course Projects
September 28, Course ProjectsSeptember 28, Course Projects
September 28, Course Projects
 
Machine Learning in Healthcare Diagnostics
Machine Learning in Healthcare DiagnosticsMachine Learning in Healthcare Diagnostics
Machine Learning in Healthcare Diagnostics
 
Building the Pacific Research Platform: Supernetworks for Big Data Science
Building the Pacific Research Platform: Supernetworks for Big Data ScienceBuilding the Pacific Research Platform: Supernetworks for Big Data Science
Building the Pacific Research Platform: Supernetworks for Big Data Science
 
"Methods for Understanding How Deep Neural Networks Work," a Presentation fro...
"Methods for Understanding How Deep Neural Networks Work," a Presentation fro..."Methods for Understanding How Deep Neural Networks Work," a Presentation fro...
"Methods for Understanding How Deep Neural Networks Work," a Presentation fro...
 
Collins seattle-2014-final
Collins seattle-2014-finalCollins seattle-2014-final
Collins seattle-2014-final
 
Mind mapping and Its Applications, Introduction to Context Trees
Mind mapping and Its Applications, Introduction to Context TreesMind mapping and Its Applications, Introduction to Context Trees
Mind mapping and Its Applications, Introduction to Context Trees
 
Network Science: Theory, Modeling and Applications
Network Science: Theory, Modeling and ApplicationsNetwork Science: Theory, Modeling and Applications
Network Science: Theory, Modeling and Applications
 
Metron seas collaboration
Metron seas collaborationMetron seas collaboration
Metron seas collaboration
 
A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...
A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...
A Distributed Deep Learning Approach for the Mitosis Detection from Big Medic...
 
Quantum computing and machine learning overview
Quantum computing and machine learning overviewQuantum computing and machine learning overview
Quantum computing and machine learning overview
 
A New Approach to Video Compression using 3D-DCT
A New Approach to Video Compression using 3D-DCTA New Approach to Video Compression using 3D-DCT
A New Approach to Video Compression using 3D-DCT
 
Abstract imakawa
Abstract imakawaAbstract imakawa
Abstract imakawa
 

En vedette

LeanKit Webinar: Managing Complex Workflows
LeanKit Webinar: Managing Complex WorkflowsLeanKit Webinar: Managing Complex Workflows
LeanKit Webinar: Managing Complex WorkflowsLeanKit
 
Digital Strategy – Closing the Gap between Your Current Digital Reality and w...
Digital Strategy – Closing the Gap between Your Current Digital Reality and w...Digital Strategy – Closing the Gap between Your Current Digital Reality and w...
Digital Strategy – Closing the Gap between Your Current Digital Reality and w...Formicio
 
Global Commodity Update
Global Commodity UpdateGlobal Commodity Update
Global Commodity UpdateAgrud
 
Folheto viagem reforco1
Folheto viagem reforco1Folheto viagem reforco1
Folheto viagem reforco1Maria Mendes
 
Mtkbab3 dan4ratna
Mtkbab3 dan4ratnaMtkbab3 dan4ratna
Mtkbab3 dan4ratnaratnatoana
 
Folheto viagem reforco
Folheto viagem reforcoFolheto viagem reforco
Folheto viagem reforcoMaria Mendes
 

En vedette (12)

ANTONE_newresume
ANTONE_newresumeANTONE_newresume
ANTONE_newresume
 
LeanKit Webinar: Managing Complex Workflows
LeanKit Webinar: Managing Complex WorkflowsLeanKit Webinar: Managing Complex Workflows
LeanKit Webinar: Managing Complex Workflows
 
Digital Strategy – Closing the Gap between Your Current Digital Reality and w...
Digital Strategy – Closing the Gap between Your Current Digital Reality and w...Digital Strategy – Closing the Gap between Your Current Digital Reality and w...
Digital Strategy – Closing the Gap between Your Current Digital Reality and w...
 
Global Commodity Update
Global Commodity UpdateGlobal Commodity Update
Global Commodity Update
 
Folheto viagem reforco1
Folheto viagem reforco1Folheto viagem reforco1
Folheto viagem reforco1
 
Isu Romania project
Isu Romania project Isu Romania project
Isu Romania project
 
Well test
Well testWell test
Well test
 
Mtkbab3 dan4ratna
Mtkbab3 dan4ratnaMtkbab3 dan4ratna
Mtkbab3 dan4ratna
 
Folheto viagem reforco
Folheto viagem reforcoFolheto viagem reforco
Folheto viagem reforco
 
Negocis europaest 270408
Negocis europaest 270408Negocis europaest 270408
Negocis europaest 270408
 
Impress top 5
Impress top 5Impress top 5
Impress top 5
 
Module 4 1
Module 4 1Module 4 1
Module 4 1
 

Similaire à REU_2015_poster_DustynTubbs_PatrickStreifel

Image Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural NetworkImage Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural NetworkAIRCC Publishing Corporation
 
Image Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural NetworkImage Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural NetworkAIRCC Publishing Corporation
 
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION cscpconf
 
IRJET- Automated Detection of Gender from Face Images
IRJET-  	  Automated Detection of Gender from Face ImagesIRJET-  	  Automated Detection of Gender from Face Images
IRJET- Automated Detection of Gender from Face ImagesIRJET Journal
 
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATIONMULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATIONijaia
 
Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
Model-Agnostic Meta-Learning for Fast Adaptation of Deep NetworksModel-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
Model-Agnostic Meta-Learning for Fast Adaptation of Deep NetworksYoonho Lee
 
SENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELS
SENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELSSENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELS
SENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELSIJDKP
 
Course Title CS591-Advance Artificial Intelligence
Course Title CS591-Advance Artificial Intelligence           Course Title CS591-Advance Artificial Intelligence
Course Title CS591-Advance Artificial Intelligence CruzIbarra161
 
Image Captioning Generator using Deep Machine Learning
Image Captioning Generator using Deep Machine LearningImage Captioning Generator using Deep Machine Learning
Image Captioning Generator using Deep Machine Learningijtsrd
 
Paper_3.pdf
Paper_3.pdfPaper_3.pdf
Paper_3.pdfChauVVan
 
A convolutional neural network framework for classifying inappropriate online...
A convolutional neural network framework for classifying inappropriate online...A convolutional neural network framework for classifying inappropriate online...
A convolutional neural network framework for classifying inappropriate online...IAESIJAI
 
DLD meetup 2017, Efficient Deep Learning
DLD meetup 2017, Efficient Deep LearningDLD meetup 2017, Efficient Deep Learning
DLD meetup 2017, Efficient Deep LearningBrodmann17
 
Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...
Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...
Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...CSCJournals
 
IRJET- A Survey on Medical Image Interpretation for Predicting Pneumonia
IRJET- A Survey on Medical Image Interpretation for Predicting PneumoniaIRJET- A Survey on Medical Image Interpretation for Predicting Pneumonia
IRJET- A Survey on Medical Image Interpretation for Predicting PneumoniaIRJET Journal
 
Deep Learning Neural Networks in the Cloud
Deep Learning Neural Networks in the CloudDeep Learning Neural Networks in the Cloud
Deep Learning Neural Networks in the CloudIJAEMSJORNAL
 
improving Profile detection using Deep Learning
improving Profile detection using Deep Learningimproving Profile detection using Deep Learning
improving Profile detection using Deep LearningSahil Kaw
 

Similaire à REU_2015_poster_DustynTubbs_PatrickStreifel (20)

Image Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural NetworkImage Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural Network
 
Image Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural NetworkImage Segmentation and Classification using Neural Network
Image Segmentation and Classification using Neural Network
 
Mini Project PPT
Mini Project PPTMini Project PPT
Mini Project PPT
 
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION
CNN FEATURES ARE ALSO GREAT AT UNSUPERVISED CLASSIFICATION
 
IRJET- Automated Detection of Gender from Face Images
IRJET-  	  Automated Detection of Gender from Face ImagesIRJET-  	  Automated Detection of Gender from Face Images
IRJET- Automated Detection of Gender from Face Images
 
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATIONMULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION
MULTI-LEVEL FEATURE FUSION BASED TRANSFER LEARNING FOR PERSON RE-IDENTIFICATION
 
Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
Model-Agnostic Meta-Learning for Fast Adaptation of Deep NetworksModel-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
 
SENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELS
SENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELSSENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELS
SENTIMENT ANALYSIS FOR MOVIES REVIEWS DATASET USING DEEP LEARNING MODELS
 
Course Title CS591-Advance Artificial Intelligence
Course Title CS591-Advance Artificial Intelligence           Course Title CS591-Advance Artificial Intelligence
Course Title CS591-Advance Artificial Intelligence
 
Image Captioning Generator using Deep Machine Learning
Image Captioning Generator using Deep Machine LearningImage Captioning Generator using Deep Machine Learning
Image Captioning Generator using Deep Machine Learning
 
Paper_3.pdf
Paper_3.pdfPaper_3.pdf
Paper_3.pdf
 
A convolutional neural network framework for classifying inappropriate online...
A convolutional neural network framework for classifying inappropriate online...A convolutional neural network framework for classifying inappropriate online...
A convolutional neural network framework for classifying inappropriate online...
 
ppt.pdf
ppt.pdfppt.pdf
ppt.pdf
 
LatentCross.pdf
LatentCross.pdfLatentCross.pdf
LatentCross.pdf
 
DLD meetup 2017, Efficient Deep Learning
DLD meetup 2017, Efficient Deep LearningDLD meetup 2017, Efficient Deep Learning
DLD meetup 2017, Efficient Deep Learning
 
Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...
Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...
Semantic Concept Detection in Video Using Hybrid Model of CNN and SVM Classif...
 
IRJET- A Survey on Medical Image Interpretation for Predicting Pneumonia
IRJET- A Survey on Medical Image Interpretation for Predicting PneumoniaIRJET- A Survey on Medical Image Interpretation for Predicting Pneumonia
IRJET- A Survey on Medical Image Interpretation for Predicting Pneumonia
 
Deep Learning Neural Networks in the Cloud
Deep Learning Neural Networks in the CloudDeep Learning Neural Networks in the Cloud
Deep Learning Neural Networks in the Cloud
 
CV _Manoj
CV _ManojCV _Manoj
CV _Manoj
 
improving Profile detection using Deep Learning
improving Profile detection using Deep Learningimproving Profile detection using Deep Learning
improving Profile detection using Deep Learning
 

REU_2015_poster_DustynTubbs_PatrickStreifel

  • 1. Passionate on Parallel REU 2015 This REU is co-funded by the ASSURE program of the Department of Defense in partnership with the National Science Foundation REU Site Program under Award No. 1263145. Accelerating Convolution Neural Network Learning with MPI Dustyn Tubbs, Saginaw Valley State University Patrick Streifel, St. Mary’s College of Maryland Advisors: Dr. Deming Chen and Ashutosh Dhar Convolution Neural Networks (CNNs) are a type of feed- forward neural network specialized for classifying objects in images. CNNs guess the class of an object in an image by applying multiple convolution filters across regions of the image. To train a CNN, these filters are adjusted based on a guess’s error, so that the updated filters produces slightly more accurate classifications. This method of supervised learning typically takes place over tens of thousands of labeled images. Background Having finished data-parallelism in programming, our future work is based on acquiring training- time metrics for published architectures (such as the GoogLeNet architecture in [1]) and implementing model parallelism as another form of CNN parallelism. After this is accomplished, we intend to release a package to the Torch7 community which will let them easily perform both model and data parallelism in their research. Future Works Ever since 2012, when the first CNN to win the biannual ImageNet Challenge (a visual recognition contest) outperformed its closest competitor by 11%, CNNs have become the state of the art for object identification. They are being used today by big names like Google and Facebook to improve image searches, auto-tag people in photograph, and countless other applications. Given the rapid development in CNN research, the need to exploit parallelization techniques to allow for both larger networks and faster training is apparent. Motivation [1] Szegedy, Christian, et al. "Going deeper with convolutions." arXiv preprint arXiv:1409.4842 (2014). [2] Krizhevsky, Alex. "One weird trick for parallelizing convolutional neural networks." arXiv preprint arXiv:1404.5997 (2014). This research project could not have been accomplished without the dedicated support of our mentors Deming Chen and Ashutosh Dhar, our REU leaders and organizers Jill Peckham, Craig Ziles, and Mathew West, and the University of Illinois at Urbana-Champaign. Special thanks to the Circuits Research Group at CSL for hosting us for the duration of our project. References and Acknowledgements Data-Parallelism in Convolution Neural Networks is accomplished by performing distribution of both the network and the data across multiple nodes. Traditionally, a CNN is trained serially on a single node, which loads the entire data set (as seen to the right). Compare this structure with the below structure, which distributes both the data and the network. Instead of training on one image at a time, the CNN can train on each node-local chunk of training data and communicate the necessary changes to every network copy across the network. This is superior in that it lets the researcher quickly train the network, adjust the topology, and train again. Data-ParallelismWe used Torch7, a framework for working with CNNs. During our research, we discovered that Torch7 had no such convenient utilities for MPI parallelization. So we decided that our goal should be to implement MPI parallelization for Torch CNNs. Goal There are two paradigms for CNN parallelization: model and data parallelism. In model parallelism, the CNN architecture is divided among several workers and trained on the same batch of images. This technique allows you to train on much larger CNNs. In data parallelism, identical copies of the CNN are maintained by each worker and trained in parallel on separate batches of images. This allows you to train a CNN on many more images at one time. Parallelization Methods