[Paper Review] pyRecLab: A Software Library for Quick Prototyping of Recommender Systems

Gabriel Sepulveda, Vicente Domı́nguez+1 more|arXiv (Cornell University)|Jun 20, 2017

Recommender Systems and Techniques7 references4 citations

TL;DR

pyRecLab is a C++-based software library with Python bindings that enables rapid prototyping of recommender systems, combining high performance with ease of use for beginners. It supports key collaborative filtering algorithms like User/Item KNN, Slope One, and FunkSVD, achieving MAE and RMSE results comparable to LibRec while offering faster inference times, particularly for testing.

ABSTRACT

This paper introduces pyRecLab, a software library written in C++ with Python bindings which allows to quickly train, test and develop recommender systems. Although there are several software libraries for this purpose, only a few let developers to get quickly started with the most traditional methods, permitting them to try different parameters and approach several tasks without a significant loss of performance. Among the few libraries that have all these features, they are available in languages such as Java, Scala or C#, what is a disadvantage for less experienced programmers more used to the popular Python programming language. In this article we introduce details of pyRecLab, showing as well performance analysis in terms of error metrics (MAE and RMSE) and train/test time. We benchmark it against the popular Java-based library LibRec, showing similar results. We expect programmers with little experience and people interested in quickly prototyping recommender systems to be benefited from pyRecLab.

Motivation & Objective

• To address the difficulty beginners face when using Java- or C#-based recommender libraries by providing a Python-friendly alternative.
• To enable quick experimentation with standard collaborative filtering methods such as User KNN, Item KNN, Slope One, and FunkSVD.
• To deliver performance comparable to mature libraries like LibRec while maintaining low entry barriers for novice programmers.
• To support flexible data input/output and efficient recommendation generation for specific users, including list-based recommendations.

Proposed method

• The library is implemented in C++ with Python/C API bindings to ensure high performance and seamless integration with Python.
• It includes a File I/O module supporting multiple formats (CSV, TSV) and customizable column mapping for input data.
• Data handlers abstract raw data into uniform structures, enabling consistent access across algorithms regardless of input format.
• Core recommendation algorithms—such as Item Average, Slope One, User/Item KNN, and FunkSVD—are implemented with parameter tuning support.
• The Python interface allows direct instantiation of algorithm objects from Python, enabling rapid prototyping and experimentation.
• Sparse matrix handling is supported to efficiently manage medium to large datasets, avoiding performance bottlenecks.

Experimental results

Research questions

• RQ1 Can a Python-based recommender library achieve performance comparable to established Java libraries like LibRec?
• RQ2 Does providing a Python-friendly interface improve accessibility for beginners and educators in recommender systems?
• RQ3 How do pyRecLab's error metrics (MAE, RMSE) compare to those of LibRec across standard benchmark datasets?
• RQ4 What is the relative training and inference speed of pyRecLab compared to LibRec, especially under varying hyperparameter settings?

Key findings

• pyRecLab achieves MAE and RMSE values on the MovieLens 100K dataset that are nearly identical to those of LibRec, with differences below 0.001 in most cases.
• For the FunkSVD algorithm, pyRecLab demonstrates faster test-time inference than LibRec across different numbers of latent factors.
• The training times of pyRecLab and LibRec are comparable, indicating that pyRecLab maintains strong performance during model learning.
• pyRecLab supports efficient recommendation generation for specific users, including list-based recommendations, a feature missing in some competing libraries.
• The library’s use of C++ with Python bindings enables high performance while retaining the ease of use expected in Python-based development environments.