Flow mapping and data distribution on mesh-based deep learning accelerator

Mirmahaleh, Seyedeh Yasaman Hosseini; Reshadi, Midia; Shabani, Hesam; Guo, Xiaochen; Bagherzadeh, Nader

doi:10.1145/3313231.3352378

Cited by 14 publications

(8 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, Figure 8a,b show improving communication delay and energy efficiency by approximately 6.15% and 10.26% for assumption 2 compared with assumption 1 after detecting defective nodes. After detecting the defective nodes and simulating them by the cycle-accurate tool, the relationship between processing elements demonstrates reducing the total hopcouns for assumption 2 compared with assumption 1, which leads to performance improvement [19,20,39]. This section analyzed the astrocyte operations and their relationship to neuron cells and proved the dependency between cholesterol distribution and the activity of neurons.…”

Section: Simulation Resultsmentioning

confidence: 99%

“…The network size depends on the total number of tasks, which are mapped onto the nodes based on the pattern of the relationship between astrocyte, neuron, and dendrite. To simulate a mesh network, this work uses a cycle-accurate SystemC-based simulator and describes the relationship between the network nodes as the traffic tables and importing them to the simulator [19,20,39]. To deploy the simulator with a multicasting traffic feature, the researchers inspired the Noxim tool [40,41].…”

Section: Simulation Resultsmentioning

confidence: 99%

“…According to the special and successful operation of astrocyte cells to distribute cholesterol into the neurons and its phagocytosis against the waste cells, the work utilizes the astrocyte pattern to IoMT traffic distribution onto mesh network's PEs and fault detection onto them. Nevertheless, the type of network-on-chip (NoC) based communication infrastructure, traffic pattern, flow mapping and data or bus scheduling method, and clustering the network nodes affect improving or deteriorating the ideal performance [19,20,39].…”

Section: Astrocyte-flow Mapping Onto Mesh Network Nodesmentioning

confidence: 99%

“…After analyzing the Alzheimer's internal symptoms and astrocyte operations, we propose an astrocyte flow-mapping method to distribute the traffic of IoMT-based applications onto a mesh network's PEs and detect defective devices. To follow the issue, this work defines a task graph of vaccination against COVID-19 and maps it onto mesh nodes, according to the network characteristics in distributing traffic in four directions and deadlock-free routing algorithms [19][20][21][22][23]. The features create an opportunity to distribute the traffic based on astrocyte cell and phagocytosis operations and cluster the mesh nodes by inspiring the cell structure.…”

mentioning

confidence: 99%

See 3 more Smart Citations

An Astrocyte-Flow Mapping on a Mesh-Based Communication Infrastructure to Defective Neurons Phagocytosis

et al. 2021

View full text Add to dashboard Cite

In deploying the Internet of Things (IoT) and Internet of Medical Things (IoMT)-based applications and infrastructures, the researchers faced many sensors and their output’s values, which have transferred between service requesters and servers. Some case studies addressed the different methods and technologies, including machine learning algorithms, deep learning accelerators, Processing-In-Memory (PIM), and neuromorphic computing (NC) approaches to support the data processing complexity and communication between IoMT nodes. With inspiring human brain structure, some researchers tackled the challenges of rising IoT- and IoMT-based applications and neural structures’ simulation. A defective device has destructive effects on the performance and cost of the applications, and their detection is challenging for a communication infrastructure with many devices. We inspired astrocyte cells to map the flow (AFM) of the Internet of Medical Things onto mesh network processing elements (PEs), and detect the defective devices based on a phagocytosis model. This study focuses on an astrocyte’s cholesterol distribution into neurons and presents an algorithm that utilizes its pattern to distribute IoMT’s dataflow and detect the defective devices. We researched Alzheimer’s symptoms to understand astrocyte and phagocytosis functions against the disease and employ the vaccination COVID-19 dataset to define a set of task graphs. The study improves total runtime and energy by approximately 60.85% and 52.38% after implementing AFM, compared with before astrocyte-flow mapping, which helps IoMT’s infrastructure developers to provide healthcare services to the requesters with minimal cost and high accuracy.

show abstract

Section: Simulation Resultsmentioning

confidence: 99%

Section: Simulation Resultsmentioning

confidence: 99%

Section: Astrocyte-flow Mapping Onto Mesh Network Nodesmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

An Astrocyte-Flow Mapping on a Mesh-Based Communication Infrastructure to Defective Neurons Phagocytosis

et al. 2021

View full text Add to dashboard Cite

show abstract

“…To improve the bandwidth and energy efficiency of the on-chip network for accelerating NoC-based DNNs, authors in [13] have proposed a method for distributing traffic in a mesh-based NoC through considering memory access mechanism in the AlexNet, VggNet, and GoogleNet trained models. Indeed, in this study, multiple distributor nodes have been designed to support multicast data transmission between PEs based on weight reuse.…”

Section: B Mesh-based Interconnection In Nn Acceleratorsmentioning

confidence: 99%

An Overview of Efficient Interconnection Networks for Deep Neural Network Accelerators

Nabavinejad

Baharloo

Chen

et al. 2020

IEEE J. Emerg. Sel. Topics Circuits Syst.

View full text Add to dashboard Cite

Deep Neural Networks (DNNs) have shown significant advantages in many domains, such as pattern recognition, prediction, and control optimization. The edge computing demand in the Internet-of-Things (IoTs) era has motivated many kinds of computing platforms to accelerate DNN operations. However, due to the massive parallel processing, the performance of the current large-scale artificial neural network is often limited by the huge communication overheads and storage requirements. As a result, efficient interconnection and data movement mechanisms for future on-chip artificial intelligence (AI) accelerators are worthy of study. Currently, a large body of research aims to find an efficient on-chip interconnection to achieve low-power and high-bandwidth DNN computing. This paper provides a comprehensive investigation of the recent advances in efficient on-chip interconnection and design methodology of the DNN accelerator design. First, we provide an overview of the different interconnection methods on the DNN accelerator. Then, the interconnection methods on the non-ASIC DNN accelerator will be discussed. On the other hand, with the flexible interconnection, the DNN accelerator can support different computing flow, which increases the computing flexibility. With this motivation, reconfigurable DNN computing with flexible on-chip interconnection will be investigated in this paper. Finally, we investigate the emerging interconnection technologies (e.g., in/near-memory processing) for the DNN accelerator design. This paper systematically investigates the interconnection networks in modern DNN accelerator designs. With this article, the readers are able to: 1) understand

show abstract