K -MLIO: Enabling K -Means for Large Data-Sets and Memory Constrained Embedded Systems

Slimani, Camelia; Stephane, Rubini; Boukhobza, Jalil

doi:10.1109/mascots.2019.00037

Cited by 4 publications

(5 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The second observation that can be drawn is that the ARIs obtained for overlapping clusters (sep = −0.2) are low as compared to mildly and well-separated ones. However, EKmeans remains comparable to K-MLIO and the low obtained ARIs are inherent to K-means clustering quality that degrades when clusters are overlapped as discussed in [3].…”

Section: B Results and Discussionmentioning

confidence: 98%

“…Otherwise, the execution continues at the maximum frequency f max to ensure that the deadline is met. (see (3) and (4) of Fig. 4).…”

Section: B Ek-means Overviewmentioning

confidence: 87%

“…Edge intelligence presents several challenges due to resource constraints of embedded devices. Memory limitations are particularly troublesome when the volume of training data cannot fully reside in memory, leading to swapping on secondary storage [3], [28]- [31]. Meanwhile, the energy limitations inherent in these devices emphasize the need for applications that are both fast and energy efficient [23].…”

Section: Introductionmentioning

confidence: 99%

“…This becomes problematic when dealing with large data sets that exceed available memory, leading to an I/O-bound execution. To address this, the K-MLIO algorithm (for K-means with Low I/O Overhead) [3] adopts a divideand-conquer approach. It divides the data set into chunks that can fit in the memory workspace, performs independent Kmeans clustering on each chunk, and combines the results.…”

Section: Introductionmentioning

confidence: 99%

“…This way, K-MLIO reads the data set only once, regardless of size, memory, and data set. It reduces the I/O cost while maintaining similar precision with the original K-means [3]. While previous studies investigated how to optimize the K-means inference and training process, to the best of our knowledge, no study focused on training a model in a predefined amount of time on resource-constrained devices on the edge.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Training K-Means on Embedded Devices: A Deadline-Aware and Energy Efficient Design

Achira,

Slimani,

Boukhobza

2023

2023 31st International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems (MASCOTS)

View full text Add to dashboard Cite

With the surge in data production, Machine Learning techniques are now commonly used to build intelligent models. Traditionally, powerful platforms process data collected from endpoint devices. However, to address security threats and minimize communication traffic, models can be learned near endpoint devices, despite their resource shortage. K-means clustering is among the most common machine learning tasks used for embedded applications. Because the system is running on scarce resources, the learning process needs to obey a certain time limit. Even if current implementations of K-means have been optimized for embedded devices, they do not consider running within a predefined time budget. In this paper, we propose a deadline-aware and energy-efficient version of K-means called Embedded K-means (EK-means) 1 , that relies on two main ideas : (1) smartly select the right subset of data to train on to meet the deadline at the expense of the smallest clustering error possible ; (2) by dropping part of the data, slack times are identified and exploited opportunistically to apply Dynamic Voltage and Frequency Scaling techniques (DVFS) so as to decrease the energy consumption of the learning task. EK-means has been built on top of an I/O optimized version of K-means for embedded devices to maintain a low I/O proportion regardless of memory constraints. EK-means allows to cluster data while meeting more than 98% of the deadlines with a loss of 1.43% of clustering quality, and an energy reduction of up to 84.26%.

show abstract

Section: B Results and Discussionmentioning

confidence: 98%

“…Otherwise, the execution continues at the maximum frequency f max to ensure that the deadline is met. (see (3) and (4) of Fig. 4).…”

Section: B Ek-means Overviewmentioning

confidence: 87%