Composite Multiclass Losses

By Robert C. Williamson, Elodie Vernet, Mark D. Reid. 2011

99 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Optimization & Training

How researchers see this paper

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

In Artificial Intelligence

In Optimization & Training

Citations in the corpus

On Structured Prediction Theory with Calibrated Convex Surrogate Losses (2017, background): “, 2010), (L,R)calibration (Ramaswamy & Agarwal, 2016), prediction calibration (Williamson et al., 2016).”
Consistent Multiclass Algorithms for Complex Performance Measures (2015, methodology): “In particular, if the CPE model η̂ is learned by a CPE algorithm that guarantees EX [‖η̂(X) − η(X)‖1]→0 as m→∞, as is done by any algorithm that minimizes a strictly proper composite multiclass loss over a suitably large function class (Vernet et al., 2011), then the above result yields ψ-consistenc”
Cross-Entropy Loss Functions: Theoretical Analysis and Applications (2023, unknown): “…& Bartlett, 2007; Liu, 2007; Dogan et al., 2016; Wang & Scott, 2020; Ramaswamy & Agarwal, 2012; Narasimhan et al., 2015; Agarwal & Agarwal, 2015; Williamson et al., 2016; Ramaswamy & Agarwal, 2016; Finocchiaro et al., 2019; Frongillo & Waggoner, 2021; Finocchiaro et al., 2022; Wang & Scott,…”
Learning Classifiers with Fenchel-Young Losses: Generalized Entropies, Margins, and Algorithms (2018, background): “This form of a loss, S Ω ( ψ − 1 ( θ ); y ), is some-times called composite (Buja et al., 2005; Reid and Williamson, 2010; Williamson et al., 2016).”
Convex Calibration Dimension for Multiclass Loss Matrices (2014, background): “Indeed, the above lemma can be shown to hold for any surrogate that is a strictly proper composite multiclass loss [29].”
Robust Bi-Tempered Logistic Loss Based on Bregman Divergences (2019, background): “Thus, (10c) is an unbiased approximate of the expected loss (9), thus is a proper loss [20].”
Learning with Complex Loss Functions and Constraints (2018, methodology): “logistic) over a sufficiently rich function class [31] (4)In our experiments, to make the best use of the available data, we use the same sample in both steps.”