Learning Bounds for Importance Weighting

By Corinna Cortes, Yishay Mansour, Mehryar Mohri. 2010

372 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistical Learning Theory

Citations in the corpus

Recommendations as Treatments: Debiasing Learning and Evaluation (2016, methodology): “We will draw upon this work, especially the learning theory of weighting approaches in (Cortes et al., 2008; 2010).”
Domain Adaptation with Conditional Transferable Components (2016, methodology): “Theoretical analysis of importance reweighting methods for correcting covariate shift has also been studied in (Cortes et al., 2010; Yu & Szepesv´ari, 2012).”
Multi-Source Domain Adaptation: A Causal View (2015, background): “Furthermore, another future line of research is to derive convergence bounds and learning guarantees for the proposed domain adaptation approaches, following (Cortes, Mansour, and Mohri 2010; Iyer, Nath, and Sarawagi 2014).”
Balanced Linear Contextual Bandits (2018, methodology): “The balancing technique is well-known in machine learning, especially in domain adaptation and studies in learning-theoretic frameworks (Huang et al. 2007), (Zadrozny 2004), (Cortes, Mansour, and Mohri 2010).”
Estimation Considerations in Contextual Bandits (2017, methodology): “The balancing technique is well-known in machine learning, especially in domain adaptation and studies in learning-theoretic frameworks [35, 65, 24].”
Optimal and Adaptive Off-policy Evaluation in Contextual Bandits (2016, methodology): “This approach was suggested by Dudík et al. [2011], and is standard practice of doubly robust estimators when the direct method (or the oracle estimator) needs to be estimated from the data.”
Representation Balancing MDPs for Off-Policy Policy Evaluation (2018, background): “The 239 constant CM n, ,t measures how complex the model class is and it is derived from traditional learning 240 theory results [4].”
Efficient Policy Learning (2017, methodology): “Meanwhile, bounds that scale as √ E [Γ2] log(n)/n are developed by Cortes et al. (2010) and Maurer and Pontil (2009); however, the additional log(n) factor makes these bounds inappropriate for asymptotic analysis.”
Policy Optimization via Importance Sampling (2018, neutral): “Importing the notation from [7], we indicate the exponentiated α-Rényi divergence as dα(P‖Q) = exp (Dα(P‖Q)).”
Stochastic Variance-Reduced Policy Gradient (2018, methodology): “Similarly to what is done in SL (Cortes et al., 2010), we require the variance of the importance weight to be bounded.”
Robust Learning under Uncertain Test Distributions: Relating Covariate Shift to Model Misspecification (2014, background): “Analyses on other forms of error were also introduced in the literature (Shimodaira, 2000; Sugiyama & Müller, 2005; Cortes et al., 2010).”
Adaptation Based on Generalized Discrepancy (2019, methodology): “Whereas there are other methods such as transfer cross validation (Zhong et al., 2010) to select the parameters for our algorithm, these methods require the use of importance weighting which as shown in (Cortes et al., 2010) is not theoretically justified.”
Characterizing and Avoiding Negative Transfer (2018, methodology): “In our approach, a discriminator estimating both marginal and joint distributions is used as a gate to filter potentially harmful source data by reducing the bias between source and target risks, which corresponds to the idea of importance reweighting [5, 37].”
Generative Adversarial Nets from a Density Ratio Estimation Perspective (2016, methodology): “A similar idea has also been used in [6].”
Batch learning from logged bandit feedback through counterfactual risk minimization (2015, background): “Other recent bounds derived from analyzing Renyi divergences (Cortes et al., 2010) can analogously be co-opted in our approach to counterfactual learning.”
Domain adaptation–can quantity compensate for quality? (2013, background): “(Cortes, Mansour, and Mohri 2010) propose a non-conservative Domain Adaptation paradigm with provable success rates, assuming different relaxations of the point-wise weight ratio assumption.”