The energy consumption of machine learning applications and their impact on the environment has recently gained attention as a research area, focusing on the model creation and training/inference phases. The data-oriented stages of the machine learning pipeline, which involve pre-processing, cleaning, and exploratory analysis, are critical components. However, energy consumption during these stages has received limited attention. Dataframe processing libraries play a significant role in these stages, and optimizing their energy consumption is important for reducing environmental impact and operational costs. Therefore, as a first step towards studying their energy efficiency, we investigate and compare the energy consumption of three popular dataframe processing libraries, namely Pandas, Vaex, and Dask. We perform experiments across 21 dataframe processing operations within four categories, utilizing three distinct datasets. Our results indicate that no single library is the most energy-efficient for all tasks, and the choice of a library can have a significant impact on energy consumption based on the types and frequencies of operations performed. The findings of this study suggest the potential for optimization of the energy consumption of data-oriented stages in the machine learning pipeline and warrant further research in this area.

Energy efficiency has become an important consideration in deep learning systems. However, it remains a largely under-emphasized aspect during the development. Despite the emergence of energy-efficient deep learning patterns, their adoption remains a challenge due to limited awareness. To address this gap, we present DENT (Deep Learning Energy Pattern Tagger, a Chrome extension used to add "energy pattern tags" to the deep learning related questions from Stack Overflow. The idea of DENT is to hint to the developers about the possible energy-saving opportunities associated with the Stack Overflow post through energy pattern labels. We hope this will increase awareness about energy patterns in deep learning and improve their adoption. A preliminary evaluation of DENT achieved an average precision of 0.74, recall of 0.66, and an F1-score of 0.65 with an accuracy of 66%. The demonstration of the tool is available at https://youtu.be/S0Wf_w0xajw and the related artifacts are available at https://rishalab.github.io/DENT/

The exponential rise of deep learning, aided by the availability of several frameworks and specialized hardware, has led to its application in a wide variety of domains. The availability of GPUs has made it easier to train networks with a huge number of parameters. However, this rise has come at the expense of ever-increasing energy requirements and carbon footprint. While the existing work tries to combat this issue by proposing optimizations in the hardware and the neural network architectures, there is an absence of general energy efficiency guidelines for deep learning developers. In this paper, we propose an initial catalog of 8 energy patterns for developing deep learning applications by analyzing 1361 posts from Stack Overflow. Our hope is that these energy patterns may help the developers adopt energy efficient practices in their deep learning projects. A survey with 14 deep learning developers showed us that the developers are largely in agreement with the usefulness of the catalog from an energy efficiency perspective. A detailed description of the catalog, along with the posts related to each energy pattern, is available at the following link: https://rishalab.github.io/dl_energy_patterns/

Energy efficiency is an essential consideration in mobile application development, given that these apps run on battery-powered devices. This has led the researchers to develop a set of energy design patterns that can help the developers improve the energy efficiency of their applications. However, the adoption of these energy patterns in projects remains a challenge, given the lack of awareness about these patterns among the developers. To bridge this gap, we propose our tool eTagger, a Google Chrome extension that tags GitHub issues from Android repositories with associated energy patterns. eTagger works based on the embeddings generated by Sentence-BERT. We believe that labeling the GitHub issues with energy patterns may help towards their larger adoption as GitHub is a prominent platform in collaborative software development. A preliminary evaluation of eTagger achieved an AUC-ROC of 0.73 with a precision of 0.58, recall of 0.53 and an F1-score of 0.5. The demonstration of the tool is available at https://youtu.be/hP4pWJ4AKxE and related artifacts at https://rishalab.github.io/eTagger/.

The availability of open source projects on platforms like GitHub has led to the wide use of the artifacts from these projects in software engineering research. These publicly available artifacts have been used to train artificial intelligence models used in various empirical studies and the development of tools. However, these advancements have missed out on the artifacts from non-open source projects due to the unavailability of the data. A major cause for the unavailability of the data from non-open source repositories is the issue concerning data privacy. In this paper, we propose using federated learning to address the issue of data privacy to enable the use of data from non-open source to train AI models used in software engineering research. We believe that this can potentially enable industries to collaborate with software engineering researchers without concerns about privacy. We present the preliminary evaluation of the use of federated learning to train a classifier to label bug-fix commits from an existing study to demonstrate its feasibility. The federated approach achieved an F1 score of 0.83 compared to a score of 0.84 using the centralized approach. We also present our vision of the potential implications of the use of federated learning in software engineering research.