Incorporating Recklessness to Collaborative Filtering based Recommender Systems
Diego Pérez-López, Fernando Ortega, Ángel González-Prieto, Jorge Dueñas-Lerín
TL;DR
This work tackles the reliability/coverage dilemma in probability-based collaborative filtering by introducing recklessness regularization, a hyperparameterized term that biases the output distribution variance to produce spikeier predictions. The authors formulate the augmented loss $\mathcal{F}_{\alpha}(P,Q)=\mathcal{F}(P,Q)-\alpha\sum_{u,i}\hat{\mathbb{V}}^{X}_{\hat{\mathbb{P}}_{ui}}$ and derive gradient updates, specializing to BeMF with softmax outputs and providing explicit updates that incorporate $\alpha$ and the derivatives of the softmax. Through experiments on FilmTrust and MovieLens (100K/1M) using NSGA-II hyper-parameter tuning, Recklessness is shown to widen the Pareto front and improve the hyper-volume, yielding better trade-offs between prediction accuracy (lower $\text{MAE}$) and coverage across reliability thresholds. The approach demonstrates that a tunable risk mechanism can enhance both the quantity and quality of recommendations, with negative $\alpha$ producing conservative, highly reliable predictions and positive $\alpha$ enabling broader, riskier predictions. This has practical impact for deploying RS with controllable risk, balancing user trust and recommendation breadth.
Abstract
Recommender systems are intrinsically tied to a reliability/coverage dilemma: The more reliable we desire the forecasts, the more conservative the decision will be and thus, the fewer items will be recommended. This causes a detriment to the predictive capability of the system, as it is only able to estimate potential interest in items for which there is a consensus in their evaluation, rather than being able to estimate potential interest in any item. In this paper, we propose the inclusion of a new term in the learning process of matrix factorization-based recommender systems, called recklessness, that takes into account the variance of the output probability distribution of the predicted ratings. In this way, gauging this recklessness measure we can force more spiky output distribution, enabling the control of the risk level desired when making decisions about the reliability of a prediction. Experimental results demonstrate that recklessness not only allows for risk regulation but also improves the quantity and quality of predictions provided by the recommender system.