ProTech Professional Technical Services, Inc.

Deep Learning With BigDL

Course Summary

Description

This course introduces concepts and BigDL library to students. The abundance of data and affordable cloud scale has led to an explosion of interest in Deep Learning. Intel has released an excellent library called BigDL to open-source, allowing state-of-the-art done at scale

Objective

 Introduction to Deep Learning  Deep Learning concepts CourseOutline  Introducing the BigDL Library  RDD API versus Pipeline API  Using BigDL with Tensorflow and models.  Visualizing Deep Learning Data with Tensorboard  Scaling Deep Learning with to massive distributed data

Topics

 Introduction to Deep Learning  Convolutional Neural Networks in BigDL  Introducing BigDL  Tensorflow and Caffe Models in BigDL  BigDL Execution Model  Recurrent Neural Networks in BigDL  Single Layer Linear  Long Short Term Memory (LSTM) in Classifier With BigDL BigDL  Hidden Layers: Intro to Deep Learning  Conclusion  High level BigDL: the Pipeline API and Dataframes

Audience

This course is designed for Developers, Data analysts, and data scientists.

Prerequisite

 Basic knowledge of Python language and Jupyter notebooks is assumed.  Basic knowledge of Linux environment would be beneficial  Some Machine Learning familiarity would be nice, but not necessary

Duration

Three Days

Due to the nature of this material, this document refers to numerous hardware and software products by their trade names. References to other companies and their products are for informational purposes only, and all trademarks are the properties of their respective companies. It is not the intent of ProTech Professional Technical Services, Inc. to use any of these names generically.

ProTech Professional Technical Services, Inc.

Deep Learning With BigDL

Course Outline

I. Introduction to Deep Learning A. Understanding Deep Learning VI. High level BigDL: the Pipeline API B. Activation Functions, Loss and Dataframes Functions, and Gradient Descent A. Using high level BigDL C. Training and Validation B. Developing a model with pipeline D. Regression vs. Classification API  Lab: Developing a pipeline API II. Introducing BigDL model A. BigDL intro B. BigDL Features VII. Convolutional Neural Networks in C. BigDL Versions BigDL

CourseOutline D. BigDL and Hadoop A. Introducing CNNs E. Apache Spark B. CNNs in BigDL  Lab: Setting up and Running C. CNN’s for image classification BigDL  Lab : Convolutional Neural Networks III. BigDL Execution Model A. Introducing BigDL’s Execution VIII. Tensorflow and Caffe Models in Model BigDL B. Understanding Basic BigDL A. How to export and import models in Layers BigDL C. Understand how BigDL runs on B. Transfer Learning with BigDL top of Apache Spark workloads. C. ImageNet and other pre-trained  Lab: BigDL Layers models.  Lab: Example with a Caffe model IV. Single Layer Linear Perceptron Classifier With BigDL IX. Recurrent Neural Networks in BigDL A. Introducing A. Introducing RNNs B. Linear Separability and Xor B. RNNs in BigDL Problem  Lab: RNN C. Activation Functions D. Softmax output X. Long Short Term Memory (LSTM) in E. Backpropagation, loss functions, BigDL and Gradient Descent A. Introducing RNNs  Lab: Single-Layer Perceptron B. RNNs in BigDL in BigDL  Lab: RNN

V. Hidden Layers: Intro to Deep XI. Conclusion Learning A. Summarize features and advantages A. Hidden Layers as a solution to XOR of BigDL problem B. Summarize Deep Learning and How B. Distributed Training with BigDL BigDL can help C. Vanishing Gradient Problem and C. Next steps ReLU D. Loss Functions E. Using Tensorboard to Visualize Training  Lab: Feedforward Neural Network Classifier in Tensorflow

Due to the nature of this material, this document refers to numerous hardware and software products by their trade names. References to other companies and their products are for informational purposes only, and all trademarks are the properties of their respective companies. It is not the intent of ProTech Professional Technical Services, Inc. to use any of these names generically.