Scale-Invariant Unconstrained Online Learning

Wojciech Kotłowski
Proceedings of the 28th International Conference on Algorithmic Learning Theory, PMLR 76:412-433, 2017.

Abstract

We consider a variant of online convex optimization in which both the instances (input vectors) and the comparator (weight vector) are unconstrained. We exploit a natural scale invariance symmetry in our unconstrained setting: the predictions of the optimal comparator are invariant under any linear transformation of the instances. Our goal is to design online algorithms which also enjoy this property, i.e. are scale-invariant. We start with the case of coordinate-wise invariance, in which the individual coordinates (features) can be arbitrarily rescaled. We give an algorithm, which achieves essentially optimal regret bound in this setup, expressed by means of a coordinate-wise scale-invariant norm of the comparator. We then study general invariance with respect to arbitrary linear transformations. We first give a negative result, showing that no algorithm can achieve a meaningful bound in terms of scale-invariant norm of the comparator in the worst case. Next, we compliment this result with a positive one, providing an algorithm which "almost" achieves the desired bound, incurring only a logarithmic overhead in terms of the norm of the instances.

Cite this Paper


BibTeX
@InProceedings{pmlr-v76-kotłowski17a, title = {Scale-Invariant Unconstrained Online Learning}, author = {Kotłowski, Wojciech}, booktitle = {Proceedings of the 28th International Conference on Algorithmic Learning Theory}, pages = {412--433}, year = {2017}, editor = {Hanneke, Steve and Reyzin, Lev}, volume = {76}, series = {Proceedings of Machine Learning Research}, month = {15--17 Oct}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v76/kotłowski17a/kotłowski17a.pdf}, url = {https://proceedings.mlr.press/v76/kot%C5%82owski17a.html}, abstract = {We consider a variant of online convex optimization in which both the instances (input vectors) and the comparator (weight vector) are unconstrained. We exploit a natural scale invariance symmetry in our unconstrained setting: the predictions of the optimal comparator are invariant under any linear transformation of the instances. Our goal is to design online algorithms which also enjoy this property, i.e. are scale-invariant. We start with the case of coordinate-wise invariance, in which the individual coordinates (features) can be arbitrarily rescaled. We give an algorithm, which achieves essentially optimal regret bound in this setup, expressed by means of a coordinate-wise scale-invariant norm of the comparator. We then study general invariance with respect to arbitrary linear transformations. We first give a negative result, showing that no algorithm can achieve a meaningful bound in terms of scale-invariant norm of the comparator in the worst case. Next, we compliment this result with a positive one, providing an algorithm which "almost" achieves the desired bound, incurring only a logarithmic overhead in terms of the norm of the instances.} }
Endnote
%0 Conference Paper %T Scale-Invariant Unconstrained Online Learning %A Wojciech Kotłowski %B Proceedings of the 28th International Conference on Algorithmic Learning Theory %C Proceedings of Machine Learning Research %D 2017 %E Steve Hanneke %E Lev Reyzin %F pmlr-v76-kotłowski17a %I PMLR %P 412--433 %U https://proceedings.mlr.press/v76/kot%C5%82owski17a.html %V 76 %X We consider a variant of online convex optimization in which both the instances (input vectors) and the comparator (weight vector) are unconstrained. We exploit a natural scale invariance symmetry in our unconstrained setting: the predictions of the optimal comparator are invariant under any linear transformation of the instances. Our goal is to design online algorithms which also enjoy this property, i.e. are scale-invariant. We start with the case of coordinate-wise invariance, in which the individual coordinates (features) can be arbitrarily rescaled. We give an algorithm, which achieves essentially optimal regret bound in this setup, expressed by means of a coordinate-wise scale-invariant norm of the comparator. We then study general invariance with respect to arbitrary linear transformations. We first give a negative result, showing that no algorithm can achieve a meaningful bound in terms of scale-invariant norm of the comparator in the worst case. Next, we compliment this result with a positive one, providing an algorithm which "almost" achieves the desired bound, incurring only a logarithmic overhead in terms of the norm of the instances.
APA
Kotłowski, W.. (2017). Scale-Invariant Unconstrained Online Learning. Proceedings of the 28th International Conference on Algorithmic Learning Theory, in Proceedings of Machine Learning Research 76:412-433 Available from https://proceedings.mlr.press/v76/kot%C5%82owski17a.html.

Related Material