Mitigating Spurious Correlations for Self-supervised Recommendation

نویسندگان

چکیده

Abstract Recent years have witnessed the great success of self-supervised learning (SSL) in recommendation systems. However, SSL recommender models are likely to suffer from spurious correlations, leading poor generalization. To mitigate existing work usually pursues ID-based or utilizes feature engineering identify features. Nevertheless, approaches sacrifice positive impact invariant features, while methods require high-cost human labeling. address problems, we aim automatically effect correlations. This objective requires 1) mask features without supervision, and 2) block negative transmission other during SSL. handle two challenges, propose an framework, which first divides user-item interactions into multiple environments with distribution shifts then learns a mechanism capture across environments. Based on mechanism, can remove for robust predictions via mask-guided augmentation. Extensive experiments datasets demonstrate effectiveness proposed framework mitigating correlations improving generalization abilities models.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Spurious Correlations in Mathematical Thinking

How does detection of correlational structure affect mathematical thinking and learning? When does correlational information lead to erroneous problem solving? Are experienced students susceptible to misleading correlations? This work attempts to answer these questions by examining a source of systematic errors termed the spurious-correlation effect. This effect is hypothesized to occur when a ...

متن کامل

Supervised Machine Learning Methods for Item Recommendation

class, 107accuracy, 40AdaBoost, 77adaptivity, 41age, 81ALS, see alternating least squaresalternating least squares, 36, 86Apache Mahout, 86area under the ROC curve, 41, 61, 82aspect model, 58association rules, 87attribute-based kNN, 81attribute-to-factor mapping, 45 59AUC, see area under the ROC curve bagging, 77, 87Bayesian Context-Aware ...

متن کامل

Spurious correlations and inference in landscape genetics.

Reliable interpretation of landscape genetic analyses depends on statistical methods that have high power to identify the correct process driving gene flow while rejecting incorrect alternative hypotheses. Little is known about statistical power and inference in individual-based landscape genetics. Our objective was to evaluate the power of causal-modelling with partial Mantel tests in individu...

متن کامل

Effects of Self-healing on Self-compassion, self-steam and Aggression of Poorly Supervised Teen Girls

Introduction: The aim of this study was to investigate the effectiveness of self-healing training on self-compassion, self-esteem, and aggression in poorly supervised teen girls. Method: The research method was quasi-experimental with a pretest-posttest-follow-up design with a control group. The research population included poorly supervised teen girls in Isfahan that were for sake of this res...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: Machine Intelligence Research

سال: 2023

ISSN: ['2731-538X', '2731-5398']

DOI: https://doi.org/10.1007/s11633-022-1374-8