Abstract image of a woman sitting on books and studying on a laptop

DeepLearning4J and Spark: Successes and Challenges - François Garillot

S
sparktc

The document discusses deep learning advancements, specifically through the deeplearning4j framework developed for Java and Scala, highlighting its commercial-grade capabilities and support for distributed training. It emphasizes applications in areas like autonomous driving and drug discovery, while addressing the challenges of training models outside major tech companies. Additionally, it mentions various tools and libraries that enhance the deep learning process, along with a call for collaboration on improving existing technologies.

Technology
1 of 17
Downloaded 13 times
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Deeplearning4J François Garillot, @huitseeker
Neural Networks & Deep Learning • graphical models w/ inputs and outputs • represents composition of differentiable functions • deep learning : expressivity exponential w.r.t depth
Interesting results • cat paper by Andrew Ng & Goole • AlexNet by Toronto • last week CNTK at speech recognition parity with humans
Industrial results • Autonomous Driving : Drive.ai, Comma.ai + the usual suspects • Drugs discovery : Deep Genomics (Frey) & Bayer • Predictive Maintenance : Thales, Bosch • optimistic pessimism (Moghimi, Manulife Financial Corp.)
DeepLearning in two steps : training, applying • training tends to require lots of data, (R) • but applying does not (embedded, etc). So that applying pre-trained models (Tensorframes) not the technical/business challenge. Enterprise : have lots of data yourself, what to apply ?
Benchmarks aren't distributed
Training, but how ? New Amazon GPU instances ?
DeepLearning4J and Spark: Successes and Challenges - François Garillot
Deep Learning Training • Facebook, Amazon, Google, Baidu, Microsoft have this distributed • But what if you’re not one of them ?
Training, but how ?
Distributing training • basically distributing SGD (R) • challenge is AllReduce Communication • Sparse updates, async communications
Deeplearning4J • the first commercial-grade, open-source, distributed deep- learning library written for Java and Scala • Skymind its commercial support arm
Scientific computing on the JVM • libnd4j : Vectorization, 32-bit addressing, linalg (BLAS!) • JavaCPP: generates JNI bindings to your CPP libs • ND4J : numpy for the JVM, native superfast arrays • Datavec : one-stop interface to an NDArray • DeepLearning4J: orchestration, backprop, layer definition • ScalNet: gateway drug, inspired from (and closely following) Keras
Reinforcement learning
Killing the bottlenecks : generic • swappable net backend : netty -> aeron (Hi Lightbend !) • better support for binary data : big indexed tables Binary, columnar, off-heap • and more (Tamiya Onodera's group @ IBM Japan): http://www.slideshare.net/ishizaki/exploiting-gpus-in-spark
And if you don't care about Deep Learning ? • Spark-6442 : better linear algebra than breeze, please. (sparse, performant, Java-compatible, and an OK license) • SystemML got a best paper at VLDB'16, how about helping out on nd4j ? • ND4J only lacks sparse, but not for long ...
Questions ?

More Related Content

PDF
Deep learning on a mixed cluster with deeplearning4j and spark
François Garillot
 
PDF
DL4J at Workday Meetup
David Kale
 
PDF
DeepLearning4J and Spark: Successes and Challenges - François Garillot
Steve Moore
 
PDF
Hands on image recognition with scala spark and deep learning4j
Guglielmo Iozzia
 
PDF
Intro to DeepLearning4J on ApacheSpark SDS DL Workshop 16
Romeo Kienzler
 
PPTX
Deep Learning: DL4J and DataVec
Josh Patterson
 
PPTX
Deep learning on Hadoop/Spark -NextML
Adam Gibson
 
PPTX
Deep learning with DL4J - Hadoop Summit 2015
Josh Patterson
 
Deep learning on a mixed cluster with deeplearning4j and spark
François Garillot
 
DL4J at Workday Meetup
David Kale
 
DeepLearning4J and Spark: Successes and Challenges - François Garillot
Steve Moore
 
Hands on image recognition with scala spark and deep learning4j
Guglielmo Iozzia
 
Intro to DeepLearning4J on ApacheSpark SDS DL Workshop 16
Romeo Kienzler
 
Deep Learning: DL4J and DataVec
Josh Patterson
 
Deep learning on Hadoop/Spark -NextML
Adam Gibson
 
Deep learning with DL4J - Hadoop Summit 2015
Josh Patterson
 

What's hot (20)

PPTX
Deep Learning on Qubole Data Platform
Shivaji Dutta
 
PDF
Snorkel: Dark Data and Machine Learning with Christopher Ré
Jen Aman
 
PPTX
Hadoop summit 2016
Adam Gibson
 
PPTX
Building Deep Learning Workflows with DL4J
Josh Patterson
 
PPTX
Amazon Deep Learning
Amanda Mackay (she/her)
 
PDF
Kaz Sato, Evangelist, Google at MLconf ATL 2016
MLconf
 
PPTX
Deep Learning Intro - Georgia Tech - CSE6242 - March 2015
Josh Patterson
 
PDF
Bringing Deep Learning into production
Paolo Platter
 
PPTX
Jake Mannix, Lead Data Engineer, Lucidworks at MLconf SEA - 5/20/16
MLconf
 
PDF
Bringing an AI Ecosystem to the Domain Expert and Enterprise AI Developer wit...
Databricks
 
PDF
Strata Beijing 2017: Jumpy, a python interface for nd4j
Adam Gibson
 
PDF
Apache SystemML - Declarative Large-Scale Machine Learning
Romeo Kienzler
 
PPTX
Boolan machine learning summit
Adam Gibson
 
PDF
Jeremy Nixon, Machine Learning Engineer, Spark Technology Center at MLconf AT...
MLconf
 
PDF
New Developments in H2O: April 2017 Edition
Sri Ambati
 
PDF
Anomaly Detection and Automatic Labeling with Deep Learning
Adam Gibson
 
PDF
Advanced deeplearning4j features
Adam Gibson
 
PPTX
Hadoop Summit 2014 - San Jose - Introduction to Deep Learning on Hadoop
Josh Patterson
 
ODP
Self driving computers active learning workflows with human interpretable ve...
Adam Gibson
 
PDF
Distributed Inference on Large Datasets Using Apache MXNet and Apache Spark ...
Databricks
 
Deep Learning on Qubole Data Platform
Shivaji Dutta
 
Snorkel: Dark Data and Machine Learning with Christopher Ré
Jen Aman
 
Hadoop summit 2016
Adam Gibson
 
Building Deep Learning Workflows with DL4J
Josh Patterson
 
Amazon Deep Learning
Amanda Mackay (she/her)
 
Kaz Sato, Evangelist, Google at MLconf ATL 2016
MLconf
 
Deep Learning Intro - Georgia Tech - CSE6242 - March 2015
Josh Patterson
 
Bringing Deep Learning into production
Paolo Platter
 
Jake Mannix, Lead Data Engineer, Lucidworks at MLconf SEA - 5/20/16
MLconf
 
Bringing an AI Ecosystem to the Domain Expert and Enterprise AI Developer wit...
Databricks
 
Strata Beijing 2017: Jumpy, a python interface for nd4j
Adam Gibson
 
Apache SystemML - Declarative Large-Scale Machine Learning
Romeo Kienzler
 
Boolan machine learning summit
Adam Gibson
 
Jeremy Nixon, Machine Learning Engineer, Spark Technology Center at MLconf AT...
MLconf
 
New Developments in H2O: April 2017 Edition
Sri Ambati
 
Anomaly Detection and Automatic Labeling with Deep Learning
Adam Gibson
 
Advanced deeplearning4j features
Adam Gibson
 
Hadoop Summit 2014 - San Jose - Introduction to Deep Learning on Hadoop
Josh Patterson
 
Self driving computers active learning workflows with human interpretable ve...
Adam Gibson
 
Distributed Inference on Large Datasets Using Apache MXNet and Apache Spark ...
Databricks
 
Ad

Viewers also liked (20)

PPTX
Sf data mining_meetup
Adam Gibson
 
ODP
Deep Learning meetup
Ivan Goloskokovic
 
PPTX
Ersatz meetup - DeepLearning4j Demo
Adam Gibson
 
PPS
¡Ése es Dios!
MJSUruguay
 
PPS
Si Alguna Vez Piensas Que Te Falle
oscar fernando
 
PPS
Debo Ser
MJSUruguay
 
PDF
Holden Karau - Spark ML for Custom Models
sparktc
 
PPTX
Brief introduction to Distributed Deep Learning
Adam Gibson
 
PPT
Pascua de Resurrección
Parroquialainmaculada
 
PPT
Jesucristo ha resucitado, el vive!
san gines
 
PDF
Accelerating Hadoop, Spark, and Memcached with HPC Technologies
inside-BigData.com
 
PPTX
Recurrent nets and sensors
Adam Gibson
 
ODP
Hubba Deep Learning
Ivan Goloskokovic
 
PDF
Skymind 深度学习 - T11 Summit
Shu Wei Goh
 
PDF
DeepLearning4J: Open Source Neural Net Platform
Turi, Inc.
 
PDF
DeepLearning4J and Spark: Successes and Challenges - François Garillot
sparktc
 
PDF
Deep Learning on Production with Spark
Shu Wei Goh
 
PDF
Deeplearning on Hadoop @OSCON 2014
Adam Gibson
 
PDF
Skymind's Platform - CN
Shu Wei Goh
 
PPTX
Deep Learning using Spark and DL4J for fun and profit
DataWorks Summit/Hadoop Summit
 
Sf data mining_meetup
Adam Gibson
 
Deep Learning meetup
Ivan Goloskokovic
 
Ersatz meetup - DeepLearning4j Demo
Adam Gibson
 
¡Ése es Dios!
MJSUruguay
 
Si Alguna Vez Piensas Que Te Falle
oscar fernando
 
Debo Ser
MJSUruguay
 
Holden Karau - Spark ML for Custom Models
sparktc
 
Brief introduction to Distributed Deep Learning
Adam Gibson
 
Pascua de Resurrección
Parroquialainmaculada
 
Jesucristo ha resucitado, el vive!
san gines
 
Accelerating Hadoop, Spark, and Memcached with HPC Technologies
inside-BigData.com
 
Recurrent nets and sensors
Adam Gibson
 
Hubba Deep Learning
Ivan Goloskokovic
 
Skymind 深度学习 - T11 Summit
Shu Wei Goh
 
DeepLearning4J: Open Source Neural Net Platform
Turi, Inc.
 
DeepLearning4J and Spark: Successes and Challenges - François Garillot
sparktc
 
Deep Learning on Production with Spark
Shu Wei Goh
 
Deeplearning on Hadoop @OSCON 2014
Adam Gibson
 
Skymind's Platform - CN
Shu Wei Goh
 
Deep Learning using Spark and DL4J for fun and profit
DataWorks Summit/Hadoop Summit
 
Ad

Similar to DeepLearning4J and Spark: Successes and Challenges - François Garillot (20)

PPTX
Demystifying Machine Learning and Artificial Intelligence
EPCC, University of Edinburgh
 
PDF
Deep Learning on Apache® Spark™: Workflows and Best Practices
Databricks
 
PDF
Deep Learning on Apache® Spark™ : Workflows and Best Practices
Jen Aman
 
PDF
Deep Learning on Apache® Spark™: Workflows and Best Practices
Jen Aman
 
PPTX
Deep Learning and Recurrent Neural Networks in the Enterprise
Josh Patterson
 
PDF
Data Science Accelerator Program
GoDataDriven
 
PPTX
Recent Advances in Machine Learning: Bringing a New Level of Intelligence to ...
Brocade
 
PDF
Distributed Models Over Distributed Data with MLflow, Pyspark, and Pandas
Databricks
 
PDF
Deep Domain
Zachary S. Brown
 
PDF
Big Data Analytics (ML, DL, AI) hands-on
Dony Riyanto
 
PDF
DeepLearning and Advanced Machine Learning on IoT
Romeo Kienzler
 
PDF
Using Crowdsourced Images to Create Image Recognition Models with Analytics Z...
Maurice Nsabimana
 
PDF
Deep learning in production with the best
Adam Gibson
 
PPT
Notes from 2016 bay area deep learning school
Niketan Pansare
 
PPTX
Deep learning with tensorflow
Charmi Chokshi
 
PPTX
Deep learning for dummies dec 23 2017
Ashok Govindarajan
 
PDF
Taming Your Deep Learning Workflow by Determined AI
desmondchanatdet
 
PDF
Introducing TensorFlow: The game changer in building "intelligent" applications
Rokesh Jankie
 
PPTX
High time to add machine learning to your information security stack
Minhaz A V
 
PDF
3 python packages
FEG
 
Demystifying Machine Learning and Artificial Intelligence
EPCC, University of Edinburgh
 
Deep Learning on Apache® Spark™: Workflows and Best Practices
Databricks
 
Deep Learning on Apache® Spark™ : Workflows and Best Practices
Jen Aman
 
Deep Learning on Apache® Spark™: Workflows and Best Practices
Jen Aman
 
Deep Learning and Recurrent Neural Networks in the Enterprise
Josh Patterson
 
Data Science Accelerator Program
GoDataDriven
 
Recent Advances in Machine Learning: Bringing a New Level of Intelligence to ...
Brocade
 
Distributed Models Over Distributed Data with MLflow, Pyspark, and Pandas
Databricks
 
Deep Domain
Zachary S. Brown
 
Big Data Analytics (ML, DL, AI) hands-on
Dony Riyanto
 
DeepLearning and Advanced Machine Learning on IoT
Romeo Kienzler
 
Using Crowdsourced Images to Create Image Recognition Models with Analytics Z...
Maurice Nsabimana
 
Deep learning in production with the best
Adam Gibson
 
Notes from 2016 bay area deep learning school
Niketan Pansare
 
Deep learning with tensorflow
Charmi Chokshi
 
Deep learning for dummies dec 23 2017
Ashok Govindarajan
 
Taming Your Deep Learning Workflow by Determined AI
desmondchanatdet
 
Introducing TensorFlow: The game changer in building "intelligent" applications
Rokesh Jankie
 
High time to add machine learning to your information security stack
Minhaz A V
 
3 python packages
FEG
 

More from sparktc (11)

PDF
Apache Spark™ Applications the Easy Way - Pierre Borckmans
sparktc
 
PPTX
Hyperparameter Optimization - Sven Hafeneger
sparktc
 
PDF
Data Science Hub & the Data Science Community - Philippe Van Impe
sparktc
 
PDF
Data Science and Beer - Kris peeters
sparktc
 
PDF
Creating an end-to-end Recommender System with Apache Spark and Elasticsearch...
sparktc
 
PPTX
Building Custom
Machine Learning Algorithms
with Apache SystemML
sparktc
 
PPTX
The Internet of Everywhere — How The Weather Company Scales
sparktc
 
PPTX
GPU Support in Spark and GPU/CPU Mixed Resource Scheduling at Production Scale
sparktc
 
PDF
STC Design - Engage
sparktc
 
PPTX
How Spark Enables the Internet of Things: Efficient Integration of Multiple ...
sparktc
 
PDF
Spark Summit EU: IBM Keynote
sparktc
 
Apache Spark™ Applications the Easy Way - Pierre Borckmans
sparktc
 
Hyperparameter Optimization - Sven Hafeneger
sparktc
 
Data Science Hub & the Data Science Community - Philippe Van Impe
sparktc
 
Data Science and Beer - Kris peeters
sparktc
 
Creating an end-to-end Recommender System with Apache Spark and Elasticsearch...
sparktc
 
Building Custom
Machine Learning Algorithms
with Apache SystemML
sparktc
 
The Internet of Everywhere — How The Weather Company Scales
sparktc
 
GPU Support in Spark and GPU/CPU Mixed Resource Scheduling at Production Scale
sparktc
 
STC Design - Engage
sparktc
 
How Spark Enables the Internet of Things: Efficient Integration of Multiple ...
sparktc
 
Spark Summit EU: IBM Keynote
sparktc
 

Recently uploaded (20)

PDF
Transform-Your-Factory-with-AI-Driven-Quality-Engineering.pdf
Kalilur Rahman
 
PDF
Transform-Your-Streaming-Platform-with-AI-Driven-Quality-Engineering.pdf
Kalilur Rahman
 
PDF
Transform-Your-Supply-Chain-with-AI-Driven-Quality-Engineering.pdf
Kalilur Rahman
 
PDF
Introduction to MCP and A2A Protocols: Enabling Agent Communication
Maxim Salnikov
 
PDF
Convolutional neural network based encoder-decoder for efficient real-time ob...
IAESIJAI
 
PDF
Advancing precision in air quality forecasting through machine learning integ...
IAESIJAI
 
PDF
INTERSPEECH 2025 「Recent Advances and Future Directions in Voice Conversion」
NU_I_TODALAB
 
DOCX
Basics of Cloud Computing - Cloud Ecosystem
suthak11
 
PDF
A hybrid framework for wild animal classification using fine-tuned DenseNet12...
IAESIJAI
 
PDF
Transform-Quality-Engineering-with-AI-A-60-Day-Blueprint-for-Digital-Success.pdf
Kalilur Rahman
 
PDF
Lung cancer patients survival prediction using outlier detection and optimize...
IAESIJAI
 
PPTX
MuleSoft-Compete-Deck for midddleware integrations
zupittejas
 
PDF
Enhancing plagiarism detection using data pre-processing and machine learning...
IAESIJAI
 
PDF
Planning-an-Audit-A-How-To-Guide-Checklist-WP.pdf
DROK2
 
PPTX
Training Program for knowledge in solar cell and solar industry
RabindranathKaibarty
 
PDF
Auditboard EB SOX Playbook 2023 edition.
DROK2
 
PDF
Data Virtualization in Action: Scaling APIs and Apps with FME
Safe Software
 
PDF
5-Ways-AI-is-Revolutionizing-Telecom-Quality-Engineering.pdf
Kalilur Rahman
 
PPTX
future_of_ai_comprehensive_20250822032121.pptx
forrandomuse090
 
PDF
LMS bot: enhanced learning management systems for improved student learning e...
IAESIJAI
 
Transform-Your-Factory-with-AI-Driven-Quality-Engineering.pdf
Kalilur Rahman
 
Transform-Your-Streaming-Platform-with-AI-Driven-Quality-Engineering.pdf
Kalilur Rahman
 
Transform-Your-Supply-Chain-with-AI-Driven-Quality-Engineering.pdf
Kalilur Rahman
 
Introduction to MCP and A2A Protocols: Enabling Agent Communication
Maxim Salnikov
 
Convolutional neural network based encoder-decoder for efficient real-time ob...
IAESIJAI
 
Advancing precision in air quality forecasting through machine learning integ...
IAESIJAI
 
INTERSPEECH 2025 「Recent Advances and Future Directions in Voice Conversion」
NU_I_TODALAB
 
Basics of Cloud Computing - Cloud Ecosystem
suthak11
 
A hybrid framework for wild animal classification using fine-tuned DenseNet12...
IAESIJAI
 
Transform-Quality-Engineering-with-AI-A-60-Day-Blueprint-for-Digital-Success.pdf
Kalilur Rahman
 
Lung cancer patients survival prediction using outlier detection and optimize...
IAESIJAI
 
MuleSoft-Compete-Deck for midddleware integrations
zupittejas
 
Enhancing plagiarism detection using data pre-processing and machine learning...
IAESIJAI
 
Planning-an-Audit-A-How-To-Guide-Checklist-WP.pdf
DROK2
 
Training Program for knowledge in solar cell and solar industry
RabindranathKaibarty
 
Auditboard EB SOX Playbook 2023 edition.
DROK2
 
Data Virtualization in Action: Scaling APIs and Apps with FME
Safe Software
 
5-Ways-AI-is-Revolutionizing-Telecom-Quality-Engineering.pdf
Kalilur Rahman
 
future_of_ai_comprehensive_20250822032121.pptx
forrandomuse090
 
LMS bot: enhanced learning management systems for improved student learning e...
IAESIJAI
 

DeepLearning4J and Spark: Successes and Challenges - François Garillot

  • 2. Neural Networks & Deep Learning • graphical models w/ inputs and outputs • represents composition of differentiable functions • deep learning : expressivity exponential w.r.t depth
  • 3. Interesting results • cat paper by Andrew Ng & Goole • AlexNet by Toronto • last week CNTK at speech recognition parity with humans
  • 4. Industrial results • Autonomous Driving : Drive.ai, Comma.ai + the usual suspects • Drugs discovery : Deep Genomics (Frey) & Bayer • Predictive Maintenance : Thales, Bosch • optimistic pessimism (Moghimi, Manulife Financial Corp.)
  • 5. DeepLearning in two steps : training, applying • training tends to require lots of data, (R) • but applying does not (embedded, etc). So that applying pre-trained models (Tensorframes) not the technical/business challenge. Enterprise : have lots of data yourself, what to apply ?
  • 7. Training, but how ? New Amazon GPU instances ?
  • 9. Deep Learning Training • Facebook, Amazon, Google, Baidu, Microsoft have this distributed • But what if you’re not one of them ?
  • 11. Distributing training • basically distributing SGD (R) • challenge is AllReduce Communication • Sparse updates, async communications
  • 12. Deeplearning4J • the first commercial-grade, open-source, distributed deep- learning library written for Java and Scala • Skymind its commercial support arm
  • 13. Scientific computing on the JVM • libnd4j : Vectorization, 32-bit addressing, linalg (BLAS!) • JavaCPP: generates JNI bindings to your CPP libs • ND4J : numpy for the JVM, native superfast arrays • Datavec : one-stop interface to an NDArray • DeepLearning4J: orchestration, backprop, layer definition • ScalNet: gateway drug, inspired from (and closely following) Keras
  • 15. Killing the bottlenecks : generic • swappable net backend : netty -> aeron (Hi Lightbend !) • better support for binary data : big indexed tables Binary, columnar, off-heap • and more (Tamiya Onodera's group @ IBM Japan): http://www.slideshare.net/ishizaki/exploiting-gpus-in-spark
  • 16. And if you don't care about Deep Learning ? • Spark-6442 : better linear algebra than breeze, please. (sparse, performant, Java-compatible, and an OK license) • SystemML got a best paper at VLDB'16, how about helping out on nd4j ? • ND4J only lacks sparse, but not for long ...