How to implement a recommender system based on one of the SVD algorithms?

Question

Algorithms

It is necessary to implement a recommender system based on one of the SVD algorithms.
There is a set of M users and N goods. There is no explicit evaluation of the goods.
The assessment can be built on the fact that the goods are sold in a certain volume and returned back for various reasons (for example, marriage).
Product pricing may change dynamically, depending on the increase/decrease in order volumes/returns.
The set of recommended products should change depending on the change in the assessment.
I studied information on algorithms for the implementation of recommender systems, partially understood what the SVD algorithm is. But I did not find a complete picture of the implementation. Can anyone help with this issue? And yet, in what form is it best to present the assessment of the product, in what range is it best to rank it? What is the best way to regulate the system? Since there is a cold start problem, then on what amount of data can you begin to trust the recommendations?

Answer 1

https://github.com/grahamjenson/list_of_recommende...

Answer 2

Not SVD, but also on the topic of Item-based collaborative filtering with your own hands