PAPERSessNet

Covers: theory of Session Information + User history
Estimated time needed: 60 minutes
Questions this item adddesses:
  • How to combine the user history and the sequence modelling into one network
How to use this item?

Read section 3 and 4 of this paper

URL: https://www.researchgate.net/publication/340721393_SESSNET_A_DEEP_HYBRID-STATE_SESSION-BASED_RECOMMENDER_SYSTEM_A_PREPRINT
Shortlist
publicShare

Sequence Based Recommendations

Omar NadaTotal time needed: ~6 hours
Learning Objectives
New era recommendations engine utilizes the sequences of data rather than just the user history. The sequence based recommender engine utilizes models from NLP to analyze user behavior within a session
Potential Use Cases
Retail Users Behavior, Next best item
Target Audience
INTERMEDIATE
Go through the following annotated items in order:
ARTICLE 1. Sequence Analysis in Recommender Engine
  • How does Amazon, Netflix, YouTube finds the next best item for a given user
3 minutes
ARTICLE 2. Introduction to Sequence Modeling Problems
  • Different types of Seqeuence Modeeling and the differences among them
6 minutes
PAPER 3. Recurrent Neural Networks with Top-k Gains for Session-basedRecommendations
  • How rNNwith top-k gains can help in ranking problem
90 minutes
PAPER 4. BERT4REC
  • How can bi-directionality help with next item recommendations
120 minutes
LIBRARY 5. BERT4REC code
  • A foundational paper that helps to understand how BERT can be used in recommendations problems
60 minutes
PAPER 6. SessNet
  • How to combine the user history and the sequence modelling into one network
60 minutes

Concepts Convered