This is the place I try to figure out how to implement recommendation algos
- movielens
- funds
- sketchFab
- lastfm
-
content based demo code
-
nearest neighbor (User-Based CF, Item-Based CF)
- issue: very inefficient to compute pairwise similarity(distance)
- demo code
-
matrix factorization (explicit/implicit ALS, SGD ...) demo
-
Learning to Rank
- LightFM coldstart
- Approximate Nearest Neighbors
- issue: can't install properyly under windows env (solved)
- annoy -- ubuntu16.04, windows server2012 (update conda to python 3.6 version)
- nmslib -- failed
- how to use it? lightFM+annoy
- issue: can't install properyly under windows env (solved)
- web-based funds recommendation, demo code
- introduction to recommendation - KNN model
- matrix factorization - ALS
- learning to rank
- content based