Billions of devices will compose the IoT system in the next few years, generating a huge amount of data. We can use fog computing to process these data, considering that there is the possibility of overloading the network towards the cloud. In this context, deep learning can treat these data, but the memory requirements of deep neural networks may prevent them from executing on a single resource-constrained device. Furthermore, their computational requirements may yield an unfeasible execution time. In this work, we propose *dn2pciot, a new algorithm to partition neural networks for efficient distributed execution. Our algorithm can optimize the neural network inference rate or the number of communications among devices. Additionally, our algorithm accounts appropriately for the shared parameters and biases of *cnn. We investigate the inference rate maximization for the LeNet model in constrained setups. We show that the partitionings offered by popular machine learning frameworks such as TensorFlow or by the general-purpose framework METIS may produce invalid partitionings for very constrained setups. The results show that our algorithm can partition LeNet for all the proposed setups, yielding up to 38% more inferences per second than METIS.
In the long term, the Internet of Things (IoT) is expected to become an integral part of people’s daily lives. In light of this technological advancement, an ever-growing number of objects with limited hardware may become connected to the Internet. In this chapter, we explore the importance of these constrained devices as well as how we can use them in conjunction with fog computing to change the future of the IoT. First, we present an overview of the concepts of constrained devices, IoT, and fog and mist computing, and then we present a classification of applications according to the amount of resources they require (e.g., processing power and memory). After that, we tie in these topics with a discussion of what can be expected in a future where constrained devices and fog computing are used to push the IoT to new limits. Lastly, we discuss some challenges and opportunities that these technologies may bring.
Billions of devices will compose the Internet of Things (IoT) in the next few years, generating a vast amount of data that will have to be processed and interpreted efficiently. Most data are currently processed on the cloud, however, this paradigm cannot be adopted to process the vast amount of data generated by the IoT, mainly due to bandwidth limits and latency requirements of many applications. Thus, we can use edge computing to process these data using the devices themselves. In this context, deep learning techniques are generally suitable to extract information from these data, but the memory requirements of deep neural networks may prevent even the inference from being executed on a single resource-constrained device. Furthermore, the computational requirements of deep neural networks may yield an unfeasible execution time. To enable the execution of neural network models on resource-constrained IoT devices, the code may be partitioned and distributed among multiple devices. Different partitioning approaches are possible, nonetheless, some of them reduce the inference rate at which the system can execute or increase the amount of communication that needs to be performed between the IoT devices. In this thesis, the objective is to distribute the inference execution of Convolutional Neural Networks to several constrained IoT devices. We proposed three automatic partitioning algorithms, which model the deep neural network as a dataflow graph and focus on the IoT features, using objective functions such as inference rate maximization or communication minimization and considering memory limitations as restrictions. The first algorithm is the Kernighan-and-Lin-based Partitioning, whose objective function is to minimize communication, respecting the memory restrictions of each device. The second algorithm is the Deep Neural Networks Partitioning for Constrained IoT Devices, which, additionally to the first algorithm, can maximize the neural network inference rate and can also account appropriately for the amount of memory required by the shared parameters and biases of Convolutional Neural Networks. Finally, the third algorithm is the Multilevel Deep Neural Networks Partitioning for Constrained IoT Devices, an algorithm that employs the multilevel approach to reduce the graph size and take advantage of the previous algorithm capabilities. The major contribution of this thesis is to show that we should consider the partitioning per neurons when partitioning deep neural networks into constrained IoT devices. When compared to the literature algorithms, communication reduction is usually the only offered objective function and there is no consideration of memory restrictions, allowing these algorithms to produce invalid partitionings. Additionally, our algorithms mostly produce better results than the approaches in the literature. Finally, another contribution is that we can use the proposed algorithms to partition into any kind of device any computation that can be expressed as a dataflow graph. 6.2 An example of (a...
Dedico este trabalho aos meus queridos avós, Celso, Miriam, Laura e Francisco; este último me acompanhou somente até metade do meu mestrado em vida, mas estará sempre em meu coração. Também dedico este trabalho à minha tia Cláudia, que pôde estar presente quase até o fim, e sempre me ajudou, motivou e deu bons conselhos, e à minha família e ao meu namorado. vii AgradecimentosAo meu orientador, Prof. Dr. Luiz Otávio Saraiva Ferreira, pela dedicação, apoio e confiança na orientação deste trabalho.À minha família, pela fé e apoio incondicional na realização deste trabalho e pela educação que recebi.Ao meu namorado, Helói, pela paciência e dedicação na revisão deste trabalho e pelo apoio em todos os momentos.Aos meus colegas de laboratório e de graduação, pelos momentos de descontração e apoio.Aos professores, funcionários e institutos que me acolheram ao longo da graduação e do mestrado.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.