Universal Rates for Active Learning

Authors: Steve Hanneke, Amin Karbasi, Shay Moran, Grigoris Velegkas

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical In this work we study the problem of actively learning binary classifiers... We provide a complete characterization of the optimal learning rates... This resolves an open question... We are now ready to state the main results of this work. Our first main result is a complete characterization of the optimal learning rates... Our next result characterizes exactly when these rates occur by specifying combinatorial complexity measures... The main body and supplementary material of the submission provide proofs for all the claims.
Researcher Affiliation Collaboration Steve Hanneke Purdue University steve.hanneke@gmail.com Amin Karbasi Yale University amin.karbasi@yale.edu Shay Moran Technion, Google Research smoran@technion.ac.il Grigoris Velegkas Yale University grigoris.velegkas@yale.edu
Pseudocode Yes Figure 1: Arbitrarily Fast Rates Algorithm (Hanneke et al., 2022). Figure 2: Exponential Rates Algorithm for Partial Classes. Figure 3: Majority Class Estimation Through Star Games. Figure 4: Exponential Rates Algorithm. Figure 5: Star Gale-Stewart Game on Infinite Sequences. Figure 6: The VCL Game (Bousquet et al., 2021). Figure 7: Sublinear Rates Algorithm. Figure 8: Active Select Algorithm (Hanneke, 2012).
Open Source Code No The answer NA means that paper does not include experiments requiring code.
Open Datasets No The answer NA means that the paper does not include experiments.
Dataset Splits No The answer NA means that the paper does not include experiments.
Hardware Specification No The answer NA means that the paper does not include experiments.
Software Dependencies No The answer NA means that the paper does not include experiments.
Experiment Setup No The answer NA means that the paper does not include experiments.