Studying the Practices of Deploying Machine Learning Projects on Docker

Openja, Moses; Majidi, Forough; Khomh, Foutse; Chembakottu, Bhagya; Li, Heng

doi:10.1145/3530019.3530039

Cited by 13 publications

(2 citation statements)

References 72 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Causal Impact Analysis [32] is a powerful statistical method designed to estimate [37] the causal effect of an intervention or treatment on an outcome of interest. This method allows [22] for the evaluation of interventions [38] using observational data. In Causal Impact Analysis, observed data are compared from a period before [5] the intervention (preintervention) [7] to the data from a period after the intervention (post-intervention).…”

Section: B Casual Impact Detection Using Machine Learning Methodsmentioning

confidence: 99%

Impact Analysis of Natural Herb on Streptozotocin-Induced Diabetic Rats using Machine Learning Method

Ray,

Susmita

2024

Preprint

View full text Add to dashboard Cite

Diabetes mellitus (DM) is a chronic metabolic disorder affecting millions globally. Effective management of DM is crucial to mitigate its complications. Natural herbs, including turmeric (Curcuma longa), aloe vera, and basil (Ocimum basilicum), are known for their medicinal properties and have been explored for their potential benefits in diabetes management. Objective of the paper is to evaluate the impact of the herbal treatment on Fasting Blood Glucose (FBG), Urea, Creatinine levels and Blood Urea Nitrogen (BUN) levels and employ machine learning models to predict and analyze the effectiveness of different doses of the herbal treatment. In this study, we divided our work into two phases. Firstly, we performed the experiment with adult Wistar albino rats and rats were made diabetic through intraperitoneal injection of freshly prepared streptozotocin (STZ). The diabetic rats were split into an untreated control group and treated test groups. The treated groups received varying doses of an aqueous solution of natural herbs (turmeric, aloe vera, and basil) administered orally through gavage for 21 days. In second phase, experimental data were collected and analyzed to detect the impact of natural herbs on diabetic rats. The paper proposes an algorithm that integrates the Bayesian Structural Time Series (BSTS) model with the Dynamic Time Warping (DTW) method. The Bayesian Structural Time Series (BSTS) model has been used as a baseline to determine the impact of natural herbs on specific health metrics such as Fasting Blood Glucose (FBG), Urea, Creatinine, and Blood Urea Nitrogen (BUN) levels. The DTW algorithm has been used in conjunction with the BSTS model to enhance performance. This combination of BSTS and DTW aims to provide a more comprehensive analysis of the effects of natural herbs on the health metrics of diabetic rats, highlighting the potential therapeutic benefits of these herbs in managing diabetes. Significant reduction (p=0.001) in FBG levels, Urea, Creatinine and BUN levels indicates the potential of this herbal combination in managing blood sugar levels. These findings suggest that the combined herbal preparation has a beneficial effect on the treatment of diabetes mellitus.

show abstract

Section: B Casual Impact Detection Using Machine Learning Methodsmentioning

confidence: 99%

Impact Analysis of Natural Herb on Streptozotocin-Induced Diabetic Rats using Machine Learning Method

Ray,

Susmita

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Nowadays, we are observing an increasing deployment of Deep Learning (DL)-based software systems in real-world applications, from personal banking to autonomous driving [1], [2]. Different easy-to-use Python-based frameworks are developed to help practitioners write their own DL codes, like TensorFlow/Keras [3], [4] and PyTorch [5].…”

Section: Introductionmentioning

confidence: 99%

An Empirical Study of Challenges in Converting Deep Learning Models

Openja¹,

Nikanjam²,

Yahmed³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

There is an increase in deploying Deep Learning (DL)-based software systems in real-world applications. Usually DL models are developed and trained using DL frameworks like TensorFlow and PyTorch. Each framework has its own internal mechanisms/formats to represent and train DL models (deep neural networks), and usually those formats cannot be recognized by other frameworks. Moreover, trained models are usually deployed in environments different from where they were developed. To solve the interoperability issue and make DL models compatible with different frameworks/environments, some exchange formats are introduced for DL models, like ONNX and CoreML. However, ONNX and CoreML were never empirically evaluated by the community to reveal their prediction accuracy, performance, and robustness after conversion. Poor accuracy or non-robust behavior of converted models may lead to poor quality of deployed DL-based software systems. We conduct, in this paper, the first empirical study to assess ONNX and CoreML for converting trained DL models. In our systematic approach, two popular DL frameworks, Keras and PyTorch, are used to train five widely used DL models on three popular datasets. The trained models are then converted to ONNX and CoreML and transferred to two runtime environments designated for such formats, to be evaluated. We investigate the prediction accuracy before and after conversion. Our results unveil that the prediction accuracy of converted models are at the same level of originals. The performance (time cost and memory consumption) of converted models are studied as well. The size of models are reduced after conversion, which can result in optimized DL-based software deployment. We also study the adversarial robustness of converted models to make sure about robustness of deployed DLbased software. Leveraging the state-of-the-art adversarial attack approaches, converted models are generally assessed robust at the same level of originals. However, obtained results show that CoreML models are more vulnerable to adversarial attacks compared to ONNX. The general message of our findings is that DL developers should be cautious on deployment of converted models that may 1) perform poorly while switching from one framework to another, 2) have challenges in robust deployment, or 3) run slowly, leading to poor quality of deployed DL-based software, including DL-based software maintenance tasks, like bug prediction.

show abstract