Step-by-step Deep Leaning Tutorials on Apache Spark using BigDL. The tutorials are inspired by Apache Spark examples, the Theano Tutorials and the Tensorflow tutorials.
-
Preface
- prologgue
- Organization of the tutorials
- advices and prerequisites for learners
-
Introduction to Spark basics(topic 1-4)
- introduction to What Spark is, current usage and application(provided by useful links from Spark official site)
- environment setting and install instructions
- RDD
- DataFrame
- SparkSAL
- Structured Streaming
-
Supervised Learning with BigDL(topic 6-8)
- install dependencies and set up the envrionment(imports)
- Introduction to Supervised Learning
-
Linear Regression with BigDL(topic 6)
- About batch training
- Data Generation
- Hyperperameter setup
- model creation with Linear layer
- Loss function
- Optimizer
- Execute Training
- Prediction on training data
- Model evaluation on random test data
-
Binary classification with logistic regression
- similar to structure in "Linear Regression with BigDL" but we bring the introduction to "BigDL's train_summary and validation summary API" and "how to use them to visualize the learing curve" here
-
Multiclass classification with logistic regression(topic 8)
- introduction to MNIST dataset (topic 7)
- rest is same to the structure in "Linear Regression with BigDL"
-
Overfitting and Regularization with BigDL
- What is overfitting with example
- Use regulariztion to solve overfitting
- regularization in BigDL
-
-
Nerual Networks with BigDL
- Introduction to neural networks
- install dependencies and imports
- mechanics of weight and gradient update
- Forward and backward(topic 5)
- Feedforward Neural Network(topic 9)
- RNN(topic 11)
- Bi-RNN(topic 13)
- LSTM(topic 12)
- CNN(topic 10 will include "batch normalization" here)
- batch normalization
- Auto-encoder(topic 14)
These neural network topics will have the same structure in "Binary classification with logistic regression"