• Corpus ID: 245769824

Deep Causal Reasoning for Recommendations

  title={Deep Causal Reasoning for Recommendations},
  author={Yaochen Zhu and Jing Yi and Jiayi Xie and Zhenzhong Chen},
Traditional recommender systems aim to estimate a user’s rating to an item based on observed ratings from the population. As with all observational studies, hidden confounders, which are factors that a ff ect both item exposures and user ratings, lead to a systematic bias in the estimation. Consequently, a new trend in recommender system research is to negate the influence of confounders from a causal perspective. Observing that confounders in recommendations are usually shared among items and… 

Mutually-Regularized Dual Collaborative Variational Auto-encoder for Recommendation Systems

A mutually-regularized dual collaborative variational auto-encoder (MD-CVAE) for recommendation that integrates two heterogeneous information sources, i.e., item content and user ratings, into the same principled variational framework where the weights of UAE are regularized by item content such that convergence to a non-optima due to data sparsity can be avoided.

Deep Deconfounded Content-based Tag Recommendation for UGC with Causal Intervention

A novel Monte Carlo (MC)-based estimator with bootstrap, which can achieve asymptotic unbiasedness provided that uploaders for the collected UGCs are i.i.d. samples from the population, and an evaluation strategy accordingly to unbiasedly evaluate causal tag recommenders is proposed.



Unbiased Learning for the Causal Effect of Recommendation

An unbiased learning framework for the causal effect of recommendation is proposed based on the inverse propensity scoring technique and it is demonstrated that the proposed method outperforms other biased learning methods in various settings.

Causal Inference for Recommender Systems

This work develops an algorithm that leverages classical recommendation models for causal recommendation and demonstrates that the proposed algorithm is more robust to unobserved confounders and improves recommendation.

Causal Effect Inference with Deep Latent-Variable Models

This work builds on recent advances in latent variable modeling to simultaneously estimate the unknown latent space summarizing the confounders and the causal effect and shows its method is significantly more robust than existing methods, and matches the state-of-the-art on previous benchmarks focused on individual treatment effects.

Causal Intervention for Leveraging Popularity Bias in Recommendation

A new training and inference paradigm for recommendation named Popularity-bias Deconfounding and Adjusting (PDA) is proposed, which removes the confounding popularity bias in model training and adjusts the recommendation score with desired popularity bias via causal intervention.

Causally Attentive Collaborative Filtering

This paper proposes to empower attention mechanism by the causal inference, which is a powerful tool to identify the real causal effects of attention mechanisms, and distills the causal information into the attention learning process, to minimize the distance between the traditional attention weights and the normalized ITE.

Multi-Source Causal Inference Using Control Variates

This work proposes a general algorithm to estimate causal effects from multiple data sources, where the average treatment effect (ATE) may be identifiable only in some datasets but not others, and shows theoretically that this reduces the variance of the ATE estimate.

Causal embeddings for recommendation

A new domain adaptation algorithm is proposed that learns from logged data containing outcomes from a biased recommendation policy and predicts recommendation outcomes according to random exposure, and is shown to be equivalent to learning to predict recommendation outcomes under a fully random recommendation policy.

Modeling Dynamic Missingness of Implicit Feedback for Sequential Recommendation

This work proposes a latent variable named “user intent ” to govern the temporal changes of item missingness, and a hidden Markov model to represent such a process, and extends the proposed framework to capture the dynamic preference of users, which results in a unified framework able to model different evolution patterns of user intent and user preference.

Disentangling User Interest and Conformity for Recommendation with Causal Embedding

DICE is presented, a general framework that learns representations where interest and conformity are structurally disentangled, and various backbone recommendation models could be smoothly integrated, and it is shown that DICE guarantees the robustness and interpretability of recommendation.

Recommendations as Treatments: Debiasing Learning and Evaluation

This paper provides a principled approach to handle selection biases by adapting models and estimation techniques from causal inference, which leads to unbiased performance estimators despite biased data, and to a matrix factorization method that provides substantially improved prediction performance on real-world data.