How to implement different ranking algorithms in tf-ranking framework?

Question

What is the algorithm behind tf-ranking? and can we use Lambdamart algorithm in tf-ranking .Can anyone suggest some good sources to learn these

Answer 1

TF Ranking provides a framework for you to implement your own algorithm. It helps you set up the components around the core of your model (input_fn, metrics and loss functions, etc.). But the scoring logic (aka scoring function) should be provided by the user.

You probably have already seen these, but just in case:

TF-Ranking: Scalable TensorFlow Library for Learning-to-Rank

and

Learning Groupwise Multivariate Scoring Functions Using Deep Neural Networks