Budget Learning via Bracketing

Durmus Alp Emre Acar, Aditya Gangrade, Venkatesh Saligrama
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:4109-4119, 2020.

Abstract

Conventional machine learning applications in the mobile/IoT setting transmit data to a cloud-server for predictions. Due to cost considerations (power, latency, monetary), it is desirable to minimise device-to-server transmissions. The budget learning (BL) problem poses the learner’s goal as minimising use of the cloud while suffering no discernible loss in accuracy, under the constraint that the methods employed be edge-implementable. We propose a new formulation for the BL problem via the concept of bracketings. Concretely, we propose to sandwich the cloud’s prediction, $g,$ via functions $h^-, h^+$ from a ‘simple’ class so that $h^- \le g \le h^+$ nearly always. On an instance $x$, if $h^+(x)=h^-(x)$, we leverage local processing, and bypass the cloud. We explore theoretical aspects of this formulation, providing PAC-style learnability definitions; associating the notion of budget learnability to approximability via brackets; and giving VC-theoretic analyses of their properties. We empirically validate our theory on real-world datasets, demonstrating improved performance over prior gating based methods.

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-acar20a, title = {Budget Learning via Bracketing}, author = {Acar, Durmus Alp Emre and Gangrade, Aditya and Saligrama, Venkatesh}, booktitle = {Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics}, pages = {4109--4119}, year = {2020}, editor = {Silvia Chiappa and Roberto Calandra}, volume = {108}, series = {Proceedings of Machine Learning Research}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/acar20a/acar20a.pdf}, url = { http://proceedings.mlr.press/v108/acar20a.html }, abstract = {Conventional machine learning applications in the mobile/IoT setting transmit data to a cloud-server for predictions. Due to cost considerations (power, latency, monetary), it is desirable to minimise device-to-server transmissions. The budget learning (BL) problem poses the learner’s goal as minimising use of the cloud while suffering no discernible loss in accuracy, under the constraint that the methods employed be edge-implementable. We propose a new formulation for the BL problem via the concept of bracketings. Concretely, we propose to sandwich the cloud’s prediction, $g,$ via functions $h^-, h^+$ from a ‘simple’ class so that $h^- \le g \le h^+$ nearly always. On an instance $x$, if $h^+(x)=h^-(x)$, we leverage local processing, and bypass the cloud. We explore theoretical aspects of this formulation, providing PAC-style learnability definitions; associating the notion of budget learnability to approximability via brackets; and giving VC-theoretic analyses of their properties. We empirically validate our theory on real-world datasets, demonstrating improved performance over prior gating based methods.} }
Endnote
%0 Conference Paper %T Budget Learning via Bracketing %A Durmus Alp Emre Acar %A Aditya Gangrade %A Venkatesh Saligrama %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-acar20a %I PMLR %P 4109--4119 %U http://proceedings.mlr.press/v108/acar20a.html %V 108 %X Conventional machine learning applications in the mobile/IoT setting transmit data to a cloud-server for predictions. Due to cost considerations (power, latency, monetary), it is desirable to minimise device-to-server transmissions. The budget learning (BL) problem poses the learner’s goal as minimising use of the cloud while suffering no discernible loss in accuracy, under the constraint that the methods employed be edge-implementable. We propose a new formulation for the BL problem via the concept of bracketings. Concretely, we propose to sandwich the cloud’s prediction, $g,$ via functions $h^-, h^+$ from a ‘simple’ class so that $h^- \le g \le h^+$ nearly always. On an instance $x$, if $h^+(x)=h^-(x)$, we leverage local processing, and bypass the cloud. We explore theoretical aspects of this formulation, providing PAC-style learnability definitions; associating the notion of budget learnability to approximability via brackets; and giving VC-theoretic analyses of their properties. We empirically validate our theory on real-world datasets, demonstrating improved performance over prior gating based methods.
APA
Acar, D.A.E., Gangrade, A. & Saligrama, V.. (2020). Budget Learning via Bracketing. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 108:4109-4119 Available from http://proceedings.mlr.press/v108/acar20a.html .

Related Material