Data-Centric Green AI An Exploratory Empirical Study

Abstract

With the growing availability of large-scale datasets, and the popularization of affordable storage and computational capabilities, the energy consumed by AI is becoming a growing concern. To address this issue, in recent years, studies have focused on demonstrating how AI energy efficiency can be improved by tuning the model training strategy. Nevertheless, how modifications applied to datasets can impact the energy consumption of AI is still an open question.To fill this gap, in this exploratory study, we evaluate if datacentric approaches can be utilized to improve AI energy efficiency. To achieve our goal, we conduct an empirical experiment, executed by considering 6 different AI algorithms, a dataset comprising 5,574 data points, and two dataset modifications (number of data points and number of features).Our results show evidence that, by exclusively conducting modifications on datasets, energy consumption can be drastically reduced (up to 92.16%), often at the cost of a negligible or even absent accuracy decline. As additional introductory results, we demonstrate how, by exclusively changing the algorithm used, energy savings up to two orders of magnitude can be achieved.In conclusion, this exploratory investigation empirically demonstrates the importance of applying data-centric techniques to improve AI energy efficiency. Our results call for a research agenda that focuses on data-centric techniques, to further enable and democratize Green AI.

A Systematic Review of Green AI

“…Albeit the training phase is intuitively the most energy-greedy phase, this results calls for a word of caution. From recent results (e.g., a study on data-centric Green AI [98]) the inference phase results to consume only a negligible fraction of the energy consumed in the training phase. Nevertheless, given the high execution rate of the inference phase, how the energy consumed by the infrequent execution of the training phase compares to one of the highly executed inference phase is still an open question.…”

“…Data-centric approaches for Green AI show that feature selection and subsampling techniques can significantly reduce the energy consumption of training machine learning models [98]. Subsampling strategies can be more sophisticated by removing data points that are expected to be redundant in terms of knowledge acquisition [35].…”

“…Out of all Green AI strategies, among the ones which report concrete saving percentages, a technique based on structure simplification for deep neural networks results to save more energy, amounting to 115% energy savings [114]. The other techniques which result to optimize energy the most are based on quantizing the inputs of decision trees [19] (97% energy savings), using data-centric Green AI techniques [98] (92% energy savings), and leveraging efficient deployment of AI algorithms via virtualized cloud fog networks (91% energy savings) [112]. Overall, more than half of the papers explicitly reporting energy saving percentages report a saving of at least 50% (⊲ 17 out of 27 papers), while only a minor number savings between 13% and 49%.…”

A systematic review of Green AI

“…Data-centric approaches for Green AI show that feature selection and subsampling techniques can significantly reduce the energy consumption of training machine learning models (Verdecchia et al, 2022b). Subsampling strategies can be more sophisticated by removing data points that are expected to be redundant in terms of knowledge acquisition (Dhabe et al, 2021).…”

“…Out of all Green AI strategies, among the ones which report concrete saving percentages, a technique based on structure simplification for deep neural networks results to save more energy, amounting to 115% energy savings (Zhang et al, 2018a). The other techniques which result to optimize energy the most are based on quantizing the inputs of decision trees (Abreu et al, 2020) (97% energy savings), using data-centric Green AI techniques (Verdecchia et al, 2022b) (92% energy savings), and leveraging efficient deployment of AI algorithms via virtualized cloud fog networks (91% energy savings) (Yosuf et al, 2021). Overall, more…”