Recommender systems are the backbone of many businesses from Netflix to Twitter. In this advanced course, you will learn about Recommender Systems and how to train them. This includes working with large datasets, feature selection, and a deeper discussion of Wide & Deep Models. Implementations will be using Tensorflow and Tensorboard to monitor the computational graph.
Target audience/Requirements:
This workshop is designed for data scientists and software developers with a basic understanding of ML/DL who would like to learn more about recommender systems. Familiarity with Python, linear algebra, neural networks and Tensorflow is advised. Attendees are expected to bring their own laptops for hands-on practical work.
- Learning about how to leverage ML to determine the most relevant information for a particular individual
- How (and why) to implement a Wide & Deep model
- Learn about Embeddings and why we use them in Recommender Systems
- Working with Tensorboard to understand, visualize and debug Tensorflow models
- Background to Recommender Systems
- [Theory] Background: Recommender systems
- [Theory] Different Approaches to Recommender Systems
- [Theory] Feature Selection
- [Hands-on] Autoencoders Meet Collaborative Filtering
- Wide & Deep Models
- [Theory] Key concepts: The Wide Component
- [Hands-on] The Wide Model: Linear Model with Crossed Feature Columns
- [Theory] Key concepts: The Deep Component
- [Hands-on] The Deep Model: Neural Network with Embeddings
- [Hands-on] Combining Wide and Deep Models
- [Hands-on] Training and Evaluating The Model
- [Theory] Introduction to Tensorboard
- [Hands-on] Tensorboard: save the computation graph to a TensorBoard summary
- Word Embeddings for Recommendation Systems
- [Theory] Why learn word embeddings?
- [Theory] The Skip-gram Model
- [Hands-on] Visualization of Embeddings
- Recap of key takeaways of recommender systems
- Resources to continue learning