Learning and Data Selection in Big Datasets
Hossein Shokri Ghadikolaei, Hadi Ghauch, Carlo Fischione, Mikael Skoglund
Abstract
Finding a dataset of minimal cardinality to characterize the optimal parameters of a model is of paramount importance in machine learning and distributed optimization over a network. This paper investigates the compressibility of large datasets. More specifically, we propose a framework that jointly learns the input-output mapping as well as the most representative samples of the dataset (sufficient dataset). Our analytical results show that the cardinality of the sufficient dataset increases sub-linearly with respect to the original dataset size. Numerical evaluations of real datasets reveal a large compressibility, up to 95%, without a noticeable drop in the learnability performance, measured by the generalization error.
BibTeX
@InProceedings{pmlr-v97-ghadikolaei19a,
title = {Learning and Data Selection in Big Datasets},
author = {Ghadikolaei, Hossein Shokri and Ghauch, Hadi and Fischione, Carlo and Skoglund, Mikael},
booktitle = {Proceedings of the 36th International Conference on Machine Learning},
pages = {2191--2200},
year = {2019},
editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan},
volume = {97},
series = {Proceedings of Machine Learning Research},
month = {09--15 Jun},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v97/ghadikolaei19a/ghadikolaei19a.pdf},
url = {https://proceedings.mlr.press/v97/ghadikolaei19a.html},
abstract = {Finding a dataset of minimal cardinality to characterize the optimal parameters of a model is of paramount importance in machine learning and distributed optimization over a network. This paper investigates the compressibility of large datasets. More specifically, we propose a framework that jointly learns the input-output mapping as well as the most representative samples of the dataset (sufficient dataset). Our analytical results show that the cardinality of the sufficient dataset increases sub-linearly with respect to the original dataset size. Numerical evaluations of real datasets reveal a large compressibility, up to 95%, without a noticeable drop in the learnability performance, measured by the generalization error.}
}