Abstract:We propose Top-N-Rank, a novel family of list-wise Learning-to-Rank models for reliably recommending the N top-ranked items. The proposed models optimize a variant of the widely used discounted cumulative gain (DCG) objective function which differs from DCG in two important aspects: (i) It limits the evaluation of DCG only on the top N items in the ranked lists, thereby eliminating the impact of low-ranked items on the learned ranking function; and (ii) it incorporates weights that allow the model to leverage multiple types of implicit feedback with differing levels of reliability or trustworthiness. Because the resulting objective function is non-smooth and hence challenging to optimize, we consider two smooth approximations of the objective function, using the traditional sigmoid function and the rectified linear unit (ReLU). We propose a family of learning-to-rank algorithms (Top-N-Rank) that work with any smooth objective function. Then, a more efficient variant, <a class="link-external link-http" href="http://Top-N-Rank.ReLU" rel="external noopener nofollow">this http URL</a>, is introduced, which effectively exploits the properties of ReLU function to reduce the computational complexity of Top-N-Rank from quadratic to linear in the average number of items rated by users. The results of our experiments using two widely used benchmarks, namely, the MovieLens data set and the Amazon Video Games data set demonstrate that: (i) The `top-N truncation' of the objective function substantially improves the ranking quality of the top N recommendations; (ii) using the ReLU for smoothing the objective function yields significant improvement in both ranking quality as well as runtime as compared to using the sigmoid; and (iii) <a class="link-external link-http" href="http://Top-N-Rank.ReLU" rel="external noopener nofollow">this http URL</a> substantially outperforms the well-performing list-wise ranking methods in terms of ranking quality.

LorSLIM: Low Rank Sparse Linear Methods for Top-N Recommendations.

Social Recommendation With Local Low Rank Matrix Approximation.

A two-step Recommendation Algorithm via Iterative Local Least Squares

Collaborative Filtering with Social Local Models

ILCR: Item-based Latent Factors for Sparse Collaborative Retrieval

LoRA-NCL: Neighborhood-Enriched Contrastive Learning with Low-Rank Dimensionality Reduction for Graph Collaborative Filtering

Low-Rank and Sparse Matrix Factorization for Scientific Paper Recommendation in Heterogeneous Network

Social Recommendation Using Low-Rank Semidefinite Program.

Social network recommendation based on low-rank matrix decomposition

Low-Rank and Sparse Matrix Completion for Recommendation

Local Weighted Matrix Factorization for Top-n Recommendation with Implicit Feedback.

A Nonnegative Latent Factor Model for Large-Scale Sparse Matrices in Recommender Systems via Alternating Direction Method

ImplicitSLIM and How it Improves Embedding-based Collaborative Filtering

LatLRR for subspace clustering via reweighted Frobenius norm minimization

Low-Complexity Decentralized Recommendation System With Similarity Constraints

Simrate: Improve Collaborative Recommendation Based On Rating Graph For Sparsity

Smoothed Low Rank and Sparse Matrix Recovery by Iteratively Reweighted Least Squares Minimization

LLM-enhanced Reranking in Recommender Systems

Active Subspace: Toward Scalable Low-Rank Learning

Large-Scale Affine Matrix Rank Minimization with a Novel Nonconvex Regularizer.

Top-N-Rank: A Scalable List-wise Ranking Method for Recommender Systems