TensorFlow in Context
Download as PPTX, PDF4 likes1,589 views
TensorFlow is an open source machine learning framework developed by Google. It has several advantages over other frameworks including its flexibility to run on CPU, GPU, and in distributed environments. The document discusses key features of TensorFlow such as its Python API and compares it to other frameworks. It also explains how TensorFlow can be used for deep learning applications on large, unstructured datasets with tools like Keras and how its distributed capabilities allow training on multiple GPUs and even across clusters.
TensorFlow in Context
- 2. About me 2 ● Master of engineer degree in Bioinformatic and modelization at INSA de Lyon ● PhD in Color Formulation by Statistical Learning at UTC and BASF Coatings (EP 2887275A1: Method and system for determining a color formula) ● Data scientist at Sfeir
- 3. 1. Deep Learning 2. TensorFlow 3. TensorFlow in Context TensorFlow in Context 3
- 5. 1 Deep Learning 1.1 What is Deep Learning? 1.2 Difference between academic research and industry application 5
- 6. 1.1 What is deep learning Artificial Intelligence Machine Learning Logistic, Regression, SVM, Neural Network Deep Learning CNN, LSTM, Neural Turing Machines 6
- 7. 1.1 What is deep learning 7
- 8. 1.2 Difference between academic research and industry application 8
- 9. 1.2 Difference between academic research and industry application Academic Research Industry Application Key Point Research Application Time Investment Long term Short term Development Environment Stand alone IDE, Compilation tools, Teamwork etc Goal Interest/ publication Problem solving 9
- 10. TensorFlow2 10
- 11. 2 TensorFlow 2.1 Key features 2.2 Comparison with others deep learning libraries 11
- 12. 2.1 Key features 12 ● Open source by Google ● Python API ● Board ● Android (SDK) Mobile application https://github.com/tensorflow/tensorflow
- 13. However, TensorFlow is very slow... 2.1 Key features https://github.com/soumith/convnet-benchmarks https://www.reddit.com/r/MachineLearning/comments/48gfop/tensorflow_speed_questions/ 13
- 14. 2.2 Comparison with others deep learning libraries http://deeplearning.net/software_links/ 14
- 15. 2.2 Comparison with others deep learning libraries Name Language OS GPU Related Library Theano Python Win, Lin, Mac CUDA,Opencl Lasagne, Keras Torch Lua, C Lin, IOS, Android CUDA Caffe C++, Python, Matlab Lin, Win, Mac CUDA, Opencl TensorFlow Python Lin, Mac, Android CUDA Keras, Skflow mxnet Python, R, Julia Lin, Windows, Mac CUDA https://github.com/zer0n/deepframeworks 15
- 16. 2.2 Comparison with others deep learning libraries Why TensorFlow? 16
- 18. 3.1 What is unique about TensorFlow? 3.2 TensorFlow with Data Science Tools 3.3 TensorFlow for Big Data 3 TensroFlow in Context 18
- 19. 3.1 What is unique about TensorFlow That would be crazy if it weren't Google 19
- 20. The author list of TensorFlow: ⬡ Jeff Dean: father of MapReduce ⬡ Ian Goodfellow: main contributor of Theano/PyLearn2 ⬡ Yangqing Jia: main contributor of Caffe ⬡ and other great Google researchers and engineers. 3.1 What is unique about TensorFlow 20
- 21. 3.2 TensorFlow with Data Science Tools Why we need deep learning in Industry application besides playing Go? 21
- 22. Avoid hand-crafted features 3.2 Tensorflow with Data Science Tools 22
- 23. No free lunch: Deep learning applications are generally applied to massive unstructured data. MNIST 60k ImageNet 50 million Yelp Restaurant Photo Classification 230 k 3.2 Tensorflow with Data Science Tools 23
- 24. Most used data science languages: TensorFlow has an API in Python Python R Data Manipulation Pandas dplyr, data.table Data Visualization Matplotlib ggplot2, ggvis Machine Learning scikit-learn caret 3.2 Tensorflow with Data Science Tools 24
- 25. Deep Learning is hard: 3.2 Tensorflow with Data Science Tools 25
- 26. Deep learning library like keras, Skflow (based on TensorFlow) were developed with a focus on enabling fast experimentation. 3.2 Tensorflow with Data Science Tools 26
- 27. No free lunch: Deep learning applications are generally applied to massive unstructured data. MNIST 60k ImageNet 50 million Yelp Restaurant Photo Classification 230 k 3.3 Tensorflow for Big Data GPU makes the deep learning training possible 27
- 28. 3.3 Tensorflow for Big Data CPU vs GPU 28
- 29. Training on Multiple-GPU: ⬡ A single GTX 580 GPU has only 3GB of memory ⬡ GPU memory limits the maximum size of the networks that can be trained ⬡ Training examples may be too big to fit on on GPU 3.3 Tensorflow for Big Data 29
- 30. 1 GPU vs multiple-GPU 3.3 Tensorflow for Big Data 30
- 31. In TensorFlow, the supported device types are CPU and GPU. They are represented as strings. For example: ⬡ "/cpu:0": The CPU of your machine. ⬡ "/gpu:0": The GPU of your machine, if you have one. ⬡ "/gpu:1": The second GPU of your machine, etc. Much earier than others libraries https://www.tensorflow.org/versions/r0.7/how_tos/using_gpu/index.html 3.3 Tensorflow for Big Data 31
- 32. 3.3 Tensorflow for Big Data GPU is not enough 32
- 33. Speed Ease of use Generality Runs Everywhere 3.3 Tensorflow for Big Data 33
- 34. https://databricks.com/blog/2016/01/25/deep-learning-with-spark-and-tensorflow.html Distributed Tensor Flow on Spark is published on early 2016 3.3 Tensorflow for Big Data 34
- 35. Tensorflow in Context Name Language OS GPU Related Library Theano Python Win, Lin, Mac CUDA,Opencl Lasagne, Keras Torch Lua, C Lin, IOS, Android CUDA Caffe C++, Python, Matlab Lin, Win, Mac CUDA, Opencl Tensorflow Python Lin, Mac, Android CUDA Keras, Skflow mxnet Python, R, Julia Lin, Windows, Mac CUDA 35
- 36. Thank you. 36
- 37. 37 by Jiqiong QIU SFEIR - Copyright ©2016
Editor's Notes
- #13: Lien github