Training on the Edge: The why and the how

Kukreja, Navjot; Shilova, Alena; Beaumont, Olivier; Hückelheim, Jan; Ferrier, Nicola; Hovland, Paul; Gorman, Gerard J.

doi:10.1109/ipdpsw.2019.00148

Cited by 29 publications

(11 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, for an input image that is large enough, or a network that is deep enough, it is seen that the input image, network weights, and network activations together require more memory than available on a single node, even for a single input image (batchsize = 1 ). We previously addressed this issue in the context of neural networks (Kukreja et al, 2019b). In this paper we address the same issue for FWI.…”

Section: Fwi and Other Similar Problemsmentioning

confidence: 96%

Lossy Checkpoint Compression in Full Waveform Inversion

Kukreja¹,

Hückelheim²,

Louboutin³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. This paper proposes a new method that combines check- pointing methods with error-controlled lossy compression for large-scale high-performance Full-Waveform Inversion (FWI), an inverse problem commonly used in geophysical exploration. This combination can signif- icantly reduce data movement, allowing a reduction in run time as well as peak memory. In the Exascale computing era, frequent data transfer (e.g., memory bandwidth, PCIe bandwidth for GPUs, or network) is the performance bottleneck rather than the peak FLOPS of the processing unit. Like many other adjoint-based optimization problems, FWI is costly in terms of the number of floating-point operations, large memory foot- print during backpropagation, and data transfer overheads. Past work for adjoint methods has developed checkpointing methods that reduce the peak memory requirements during backpropagation at the cost of additional floating-point computations. Combining this traditional checkpointing with error-controlled lossy compression, we explore the three-way tradeoff between memory, precision, and time to solution. We investigate how approximation errors introduced by lossy compression of the forward solution impact the objective function gradient and final inverted solution. Empirical results from these numerical experiments indicate that high lossy-compression rates (compression factors ranging up to 100) have a relatively minor impact on convergence rates and the quality of the final solution.

show abstract

Section: Fwi and Other Similar Problemsmentioning

confidence: 96%

Lossy Checkpoint Compression in Full Waveform Inversion

Kukreja¹,

Hückelheim²,

Louboutin³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Traditionally the training of AI models works on a principle of centralized deployment architecture which carries heavy traffic in both communication directions (Kukreja et al, 2019;Zhou et al, 2019), eventually creating overhead. In our presented IPN, MLdriven federated learning technique also known as collaborative learning is used to train the AI model with an aim to port onto edge devices.…”

Section: Edge Trainingmentioning

confidence: 99%

Edge Intelligence in Private Mobile Networks for Next-Generation Railway Systems

et al. 2021

View full text Add to dashboard Cite

The integration of Private Mobile Networks (PMN) with edge intelligence is expected to play an instrumental role in realizing the next generation of industry applications. This combination collectively termed Intelligent Private Networks (IPN) deployed within the scope of specific industries such as transport systems can unlock several use cases and critical applications that in turn can address rising business demands. This article presents a conceptual IPN that hosts intelligence at the network edge employing emerging technologies that satisfy a number of Next-Generation Railway System (NGRS) applications. NGRS use cases along with their applications and respective beyond 5G (B5G) enabling technologies have been discussed along with possible future research and development directions that will allow these promising technologies to be used and implemented widely.

show abstract

“…Recently, an on-device ML paradigm is gaining attention in which the edge device itself processes data [1][2][3][4]. This in turn reduces the potential security problems when data are transmitted from the edge device to the server as well as the energy consumption generated from data communication.…”

Section: Related Workmentioning

confidence: 99%

Hardware/Software Co-Design for TinyML Voice-Recognition Application on Resource Frugal Edge Devices

Kwon

Park

2021

Applied Sciences

View full text Add to dashboard Cite

On-device artificial intelligence has attracted attention globally, and attempts to combine the internet of things and TinyML (machine learning) applications are increasing. Although most edge devices have limited resources, time and energy costs are important when running TinyML applications. In this paper, we propose a structure in which the part that preprocesses externally input data in the TinyML application is distributed to the hardware. These processes are performed using software in the microcontroller unit of an edge device. Furthermore, resistor–transistor logic, which perform not only windowing using the Hann function, but also acquire audio raw data, is added to the inter-integrated circuit sound module that collects audio data in the voice-recognition application. As a result of the experiment, the windowing function was excluded from the TinyML application of the embedded board. When the length of the hardware-implemented Hann window is 80 and the quantization degree is 2−5, the exclusion causes a decrease in the execution time of the front-end function and energy consumption by 8.06% and 3.27%, respectively.

show abstract

Training on the Edge: The why and the how

Cited by 29 publications

References 13 publications

Lossy Checkpoint Compression in Full Waveform Inversion

Lossy Checkpoint Compression in Full Waveform Inversion

Edge Intelligence in Private Mobile Networks for Next-Generation Railway Systems

Hardware/Software Co-Design for TinyML Voice-Recognition Application on Resource Frugal Edge Devices

Contact Info

Product

Resources

About