Collective Stability in Structured Prediction: Generalization from One Example
By Ben London, Bert Huang, B. Taskar, Lise Getoor. 2013
40 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.
- Ben London
- Corinna Cortes
- Yoshua Bengio
- Mehryar Mohri
- Lise Getoor
- Bert Huang
- Vitaly Kuznetsov
- Xu Sun
Citations in the corpus
- Stable and Fair Classification (2019, neutral): “, 2018) and structured prediction (London et al., 2013; 2014; 2016).”
- Hinge-Loss Markov Random Fields and Probabilistic Soft Logic (2015, background): “…(Pujara et al., 2013), extracting and evaluating natural-language arguments on the Web (Samadi et al., 2016), high-level computer vision (London et al., 2013), drug discovery (Fakhraei et al., 2014) and predicting drug-drug interactions (Sridhar et al., 2016), natural language semantics…”
- Stability and Generalization in Structured Prediction (2016, methodology): “In prior work (London et al., 2013, 2014), we used the term collective stability to describe the stability of the predictions, guaranteeing collective stability for predictors whose inference objectives are strongly convex.”
- Structure Regularization for Structured Prediction (2014, background): “On generalization risk analysis, related studies include [2, 14] on non-structured classification and [21, 8, 7] on structured classification.”
- The Benefits of Learning with Strongly Convex Approximate Inference (2015, background): “The other related line of work is from London et al. (2013; 2014), who derived PAC learning bounds for structured prediction .”
- PAC-Bayesian Collective Stability (2014, methodology): “Though uniform stability seems like a strong requirement, it is met by a broad class of models used in practice (London et al., 2013).”
- Structured Prediction Theory Based on Factor Graph Complexity (2016, background): “Existing structured prediction learning guarantees hold only for specific decomposable losses such as the Hamming loss [Taskar et al., 2003, Cortes et al., 2014, London et al., 2013, Collins, 2001] and do not cover many other natural loss functions, such as the edit-distance.”