Consistency and Finite Sample Behavior of Binary Class Probability Estimation

Authors: Alexander Mey, Marco Loog8967-8974

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We investigate to which extent one can recover class probabilities within the empirical risk minimization (ERM) paradigm. ... We then derive conditions under which this estimator will converge with high probability to the true class probabilities with respect to the L1-norm. One of our core contributions is a novel way to derive finite sample L1-convergence rates of this estimator for different surrogate loss functions.
Researcher Affiliation Academia 1 Delft University of Technology, The Netherlands 2 University of Copenhagen, Denmark
Pseudocode No No pseudocode or algorithm blocks are present in the paper.
Open Source Code No No statement about open-source code release or repository links for the described methodology is provided.
Open Datasets No The paper is theoretical and does not conduct experiments on a specific dataset. It refers to a 'finite i.i.d. sample (xi, yi)1 i n drawn from a distribution P on X Y' in a general statistical learning context, not a specific publicly available dataset for training.
Dataset Splits No The paper is theoretical and focuses on mathematical derivations; it does not describe empirical experiments or specific dataset splits for training, validation, or testing.
Hardware Specification No No hardware specifications are mentioned, as the paper focuses on theoretical derivations and does not report on computational experiments.
Software Dependencies No No software dependencies with specific version numbers are mentioned, as the paper is theoretical and does not describe computational implementations or experiments.
Experiment Setup No No experimental setup details such as hyperparameters or system-level training settings are provided, as the paper focuses on theoretical analysis rather than empirical experiments.