Mokhles A. Mohsin
ABSTRACT
Machine learning has become one of the cornerstones of information technology. Many machine learning algorithms have found their way into mobile devices, which have stringent requirements. Also, machine learning algorithms, such as classification and clustering, are becoming complex, requiring high processing power, thus affecting the speedup. In this paper, we introduce unique, novel, and efficient hardware architecture to accelerate the K-nearest neighbor classifier on mobile devices, considering constraints associated with these devices. We evaluate the efficiency of our hardware architecture, in terms of speedup, space, and accuracy. Our design is generic, parameterized, and scalable. Our hardware design achieves 127 times speedup compared to its software counterpart, and can also achieve 100% classification accuracy.
- Kotsiantis, S., "Supervised Machine Learning: A Review of Classification Techniques", in Proc. of Conf. on Emerging Artificial Intelligence Applications in Computer Engineering: Real Word AI Systems with Applications in eHealth, HCI, Information Retrieval and Pervasive Technologies, pp:3--24, July 2007. Google Scholar
Digital Library
- Larose, D., "Discovering Knowledge in Data: An Introduction to Data Mining", John Wiley & Sons, Inc. 2005. Google ScholarDigital Library
- Amanpreet, S., et al, "A Review of Supervised Machine Learning Algorithms", in Proc. of Int. Conf. on Computing for Sustainable Global Development, Mar. 2016.Google Scholar
- Pu Y., J. et al, "An efficient K-NN algorithm implemented on FPGA based heterogeneous computing system using OpenCL", in Proc. of IEEE 23rd Int. Symp. on Field-Programmable Custom Computing Machines, May 2015. Google ScholarDigital Library
- Hussain H., et al, "An adaptive FPGA implementation of multi-core K-nearest neighbor ensemble classifier using dynamic partial reconfiguration", in Proc. of 22nd Int. Conf. on Field Programmable Logic and Applications, Aug. 2012.Google Scholar
- Satyanarayana N., et al, "Survey of Classification Techniques in Data Mining", Int.. Journal of Innovative Science, Engineering & Technology, vol. 1, no. 9, Nov. 2014.Google Scholar
- Stamoulis I. and E. Manolakos, "Parallel Architectures for the K-NN Classifier -- Design of Soft IP Cores and FPGA Implementations", ACM Transactions on Embedded Computing Systems Journal, vol. 13, no. 2, Sep. 2013. Google ScholarDigital Library
- https://www.coursera.org/learn/machine-learning#ratingsGoogle Scholar
- Perera, D.G. and K.F. Li, "FPGA-Based Reconfigurable Hardware for Compute Intensive Data Mining Applications," in Proc. of IEEE Int. Conf. on P2P, Parallel, Grid, Cloud and Internet Computing (3PGCIC'11), pp. 100--108, Oct. 2011. Google ScholarDigital Library
- https://archive.ics.uci.edu/ml/datasets/IrisGoogle Scholar
- https://archive.ics.uci.edu/ml/datasets/Statlog+(Heart)Google Scholar
- https://archive.ics.uci.edu/ml/datasets/User+Knowledge+ModelingGoogle Scholar
- Li Z., et al, "K-Nearest Neighbor Algorithm Implementation on FPGA Using High Level Synthesis", in Proc. of 13th IEEE Int. Conf. on Solid-State and Integrated Circuit Technology, pp. 600--602, Oct. 2016.Google Scholar
Recommendations
Reverse Nearest Neighbour by Region on Mobile Devices
AINA '14: Proceedings of the 2014 IEEE 28th International Conference on Advanced Information Networking and ApplicationsReverse Nearest Neighbour queries is known for the heavyweight algorithm that makes it difficult to be implemented in mobile devices due to high computations needed to verify the objects. Since the rapid development of mobile devices' hardware and also ...
Modified nearest neighbour classifier for hyperspectral data classification
A modified k-nearest neighbour k-NN classifier is proposed for supervised remote sensing classification of hyperspectral data. To compare its performance in terms of classification accuracy and computational cost, k-NN and a back-propagation neural ...
A new general nearest neighbor classification based on the mutual neighborhood information
The nearest neighbor (NN) rule is effective for many applications in pattern classification, such as the famous k-nearest neighbor (kNN) classifier. However, NN-based classifiers perform a one-sided classification by finding the nearest neighbors simply ...
Comments