Abstract: Session Objective:
The classical approaches for RecSys are not enough efficient in capturing the dynamic behaviour of customer actions and purchase patterns. We propose the multiple/distributed Q table approaches which can deal with large state-action space and that aides in actualising the Q learning algorithm in the recommendation and huge state-action space.
- Gain an understanding of deep reinforcement learning-based recommendation system.
- How to train and evaluate the RL model with distributed Q-table?
- Reference architecture of recommendation engines.
- Adoption and Impact
- Data Scientist and Machine Learning Engineer, Data Engineers, Data Architects
The tutorial will be mainly divided in below sections:
1. Introduction: Recommendation Systems
2. Classical approaches for building a recommendation system.
3. Limitations of classical approaches
4. Introduction: Reinforcement learning
6. Deep reinforcement learning-based algorithm for building a recommendation system.
7. Training methodology and result discussion
8. Working demo
9. Case study discussion
Python Programming, Basic ML libraries in Python, Jupyter Notebook, Introduction to Recommendation system.
Bio: Ravi has professional experience of eight years in AI and ML at scale with expertise in building enterprise solutions and ML Engineering. He is part of the Centre of Excellence and responsible for building ML products from inception to production. He has worked on multiple engagements with clients mainly from Automobile and Retail industry across geographies.
He holds a bachelor's degree in Computer Science with a proficiency course in Reinforcement Learning from Indian Institute of Science. He is a certified Google Cloud Architect and Kubeflow contributor.