As the title suggests, what resources do you suggest to learn recommender systems ML to reach an intermediate-like level
4 Likes
In summary, I read ML articles.
I read some of the papers that are cited in papers that I read at major machine learning conferences. I also keep an eye on tech blogs for updates on new implementations and study the papers pertaining to the algorithms that were employed. I test the theory on a real-world issue after reading. Repeat after rinsing.
3 Likes
MS in Statistics here: I love this and do the same but have a follow-up.
How do you find datasets you like for these projects? For example, I want to run a GCN or GraphSAGE to create a recommendation system…but am unsure where to find a good dataset. Also a bit concerned about how to implement it because I haven’t seen too many examples outside of papers that are very high-level and focused on the model, not the deployment.
3 Likes
Here, I have an MS in Statistics. I enjoy doing this and have a follow-up.
For these kinds of initiatives, how do you locate datasets you like? To build a recommendation system, for instance, I would like to run a GCN or GraphSAGE, but I’m not sure where to look for a quality dataset. Concerning its implementation, I’m also a little worried because the few instances I’ve seen outside of high-level publications have mostly dealt with the model rather than the deployment.
2 Likes
Recommenders are mainly different in the metrics and that features are user-item aggs. Try to understand how they’re evaluated and why.
In implementation, the state of the art is custom neural network architectures e.g. two-tower. However, if you want to use the traditional models, which work well and easier, I like turicreate rank factorization.
1 Like
I think Search Engines: 11-442 / 11-642 (the book is Introduction to Information Retrieval) is a good starting point. It lays the foundation nicely for information retrieval, which I think you should start with. After that: papers and blogposts.