University of Twente Student Theses
Deploying Machine Learning Models on Resource Constrained Devices: A Case Study
Kashif, Mohammad Umar (2024) Deploying Machine Learning Models on Resource Constrained Devices: A Case Study.
|
PDF
2MB |
| Abstract: | In the modern day, Machine Learning and Artificial Intelligence Systems have grown exponentially in their capabilities of performing a wide range of tasks, but with that so has their energy demands in the training phase of development and the inference phase on the end device. This has led to severe concerns about their impact on global greenhouse gas emissions. It is not realistic to expect the new era of ML to come to a halt to address these environmental concerns, therefore there exists a need to explore ways to improve the efficiency of these ML models to consume fewer resources. This paper explores some potential improvements to this process, namely deploying Machine Learning models on resource-constrained IoT devices, reducing the amount of data needed to train these models, and minimizing the number of neurons needed to develop them. For the practical aspect of research, we will be exploring the most efficient manner of developing Machine Learning for motion classification on the cloud using Edge Impulse and deploying this model on a Thingy 52, a small IoT Device by Nordic Semiconductors. We will explore the effect of reducing the amount of training data required, number of training epochs, hidden layers, and neurons to converge on an acceptable model despite the reduced training factors and with the limiting resources of the Thingy 52, as well as discussing the various issues encountered and potential future improvements. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 01 general works, 50 technical science in general, 54 computer science, 81 education, teaching |
| Programme: | Computer Science BSc (56964) |
| Link to this item: | https://purl.utwente.nl/essays/101088 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page