Using a reconfigurable compute cluster for the acceleration of neural networks

In this paper we present the RAPTOR family as an advanced modular platform for both FPGA-based rapid prototyping and hardware acceleration. Using modern FPGAs and high speed communication links, performance and flexibility of the approach will be shown by means of Kohonens self-organizing map algorithm. This highly parallel algorithm is partitioned onto several FPGAs in different system environments, such as to demonstrate the scalability and the flexibility of the proposed platforms.

[1]  Junbai Wang,et al.  Clustering of the SOM easily reveals distinct gene expression patterns: results of a reanalysis of lymphoma study , 2002, BMC Bioinformatics.

[2]  P. Ienne,et al.  Performance of digital neuro-computers , 1994, Proceedings of the Fourth International Conference on Microelectronics for Neural Networks and Fuzzy Systems.

[3]  S. P. Lloyd,et al.  Least squares quantization in PCM , 1982, IEEE Trans. Inf. Theory.

[4]  Vijayakumar Chinnadurai,et al.  Segmentation and grading of brain tumors on apparent diffusion coefficient images using self-organizing maps , 2007, Comput. Medical Imaging Graph..

[5]  Mario Porrmann,et al.  Rapid Prototyping of Next-Generation Multiprocessor SoCs , 2009 .

[6]  Erzsébet Merényi,et al.  Mapping surface materials on Mars from Mars Pathfinder spectral images with HYPEREYE , 2004, International Conference on Information Technology: Coding and Computing, 2004. Proceedings. ITCC 2004..