RECORD
The price of bandit information in multiclass online classification
Machine Learning · Submitted 5 Feb 2013
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim We show that the ratio between the error rates in the two scenarios is at most $8\cdot|Y|\cdot \log(|Y|)$ in the realizable case, and $\tilde{O}(\sqrt{|Y|})$ in the agnostic case.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
We consider two scenarios of multiclass online learning of a hypothesis class $H\subseteq Y^X$. In the {\em full information} scenario, the learner is exposed to instances together with their labels. In the {\em bandit} scenario, the true label is not exposed, but rather an indication whether the learner's prediction is correct or not. We show that the ratio between the error rates in the two scenarios is at most $8\cdot|Y|\cdot \log(|Y|)$ in the realizable case, and $\tilde{O}(\sqrt{|Y|})$ in the agnostic case. The results are tight up to a logarithmic factor and essentially answer an open question from (Daniely et. al. - Multiclass learnability and the erm principle). We apply these results to the class of $\gamma$-margin multiclass linear classifiers in $\reals^d$. We show that the bandit error rate of this class is $\tilde{\Theta}(\frac{|Y|}{\gamma^2})$ in the realizable case and $\tilde{\Theta}(\frac{1}{\gamma}\sqrt{|Y|T})$ in the agnostic case. This resolves an open question from (Kakade et. al. - Efficient bandit algorithms for online multiclass prediction).
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | We consider two scenarios of multiclass online learning of a hypothesis class $H\subseteq Y^X$. | background 0.71 | 0.35 | 0.02 | asserted 0.68 |
| s1 | In the {\em full information} scenario, the learner is exposed to instances together with their labels. | background 0.91 | 0.08 | 0.01 | asserted 0.98 |
| s2 | In the {\em bandit} scenario, the true label is not exposed, but rather an indication whether the learner's prediction is correct or not. | background 0.87 | 0.09 | 0.02 | asserted 0.98 |
| s3 | We show that the ratio between the error rates in the two scenarios is at most $8\cdot|Y|\cdot \log(|Y|)$ in the realizable case, and $\tilde{O}(\sqrt{|Y|})$ in the agnostic case. | result 1.00 | 0.96 | 0.81 | asserted 1.00 |
| s4 | The results are tight up to a logarithmic factor and essentially answer an open question from (Daniely et. al. - Multiclass learnability and the erm principle). | interpretation 0.88 | 0.81 | 0.22 | asserted 0.99 |
| s5 | We apply these results to the class of $\gamma$-margin multiclass linear classifiers in $\reals^d$. | contribution 0.88 | 0.85 | 0.03 | asserted 0.81 |
| s6 | We show that the bandit error rate of this class is $\tilde{\Theta}(\frac{|Y|}{\gamma^2})$ in the realizable case and $\tilde{\Theta}(\frac{1}{\gamma}\sqrt{|Y|T})$ in the agnostic case. | result 1.00 | 0.96 | 0.21 | asserted 1.00 |
| s7 | This resolves an open question from (Kakade et. al. - Efficient bandit algorithms for online multiclass prediction). | interpretation 0.71 | 0.71 | 0.10 | asserted 1.00 |
What was examined, and what was not
- arXiv listing abstract v2
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
We consider two scenarios of multiclass online learning of a hypothesis class $H\subseteq Y^X$. In the {\em full information} scenario, the learner is exposed to instances together with their labels. In the {\em bandit} scenario, the true label is not exposed, but rather an indication whether the learner's prediction is correct or not. We show that the ratio between the error rates in the two scenarios is at most $8\cdot|Y|\cdot \log(|Y|)$ in the realizable case, and $\tilde{O}(\sqrt{|Y|})$ in the agnostic case. The results are tight up to a logarithmic factor and essentially answer an open question from (Daniely et. al. - Multiclass learnability and the erm principle). We apply these results to the class of $\gamma$-margin multiclass linear classifiers in $\reals^d$. We show that the bandit error rate of this class is $\tilde{\Theta}(\frac{|Y|}{\gamma^2})$ in the realizable case and $\tilde{\Theta}(\frac{1}{\gamma}\sqrt{|Y|T})$ in the agnostic case. This resolves an open question from (Kakade et. al. - Efficient bandit algorithms for online multiclass prediction).
Human review
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.