Reducing the Friction for Building Recommender Systems with Merlin

Abstract

Recommender Systems (RecSys) are the engine of the modern internet and the catalyst for human decisions. The goal of a recommender system is to generate relevant recommendations for users from a collection of items or services that might interest them. Building a recommendation system is challenging because it requires multiple stages (item retrieval, filtering, ranking, ordering) to work together seamlessly and efficiently during training and inference. The biggest challenges faced by new practitioners are the lack of understanding around what RecSys look like in the real world and the difficulty in transitioning from the simple Matrix Factorization (MF) to more complex deep learning architectures with multiple input features, neural components and prediction heads. To address these challenges on building recommender systems, NVIDIA developed an open source framework, called Merlin. Merlin consists of a set of libraries and tools to help RecSys practitioners build models and pipelines easily and more efficiently. Merlin Models provides modularized building blocks that can be easily connected to build classic and state-of-the-art models. It offers flexibility at each stage: multiple input processing/representation modules, different layers for designing the model's architecture, prediction heads, loss functions, negative sampling techniques, among others. In this hands-on tutorial, participants will start with data preparation using NVTabular an open-source feature engineering and preprocessing library designed to quickly and easily manipulate large scale datasets. Participants will then work on modeling with Merlin Models library, building the fundamental recommendation models such as MF and then transitioning to more complex deep learning-based models for candidate retrieval. In each iteration, we will demonstrate the seamless integration between data preparation and model training. Over the span of this tutorial, participants will learn the fundamentals of recommender systems modeling and how to build a two-stage recommender system easily using open source Merlin libraries.