To verify our work, we compared our implementation with traditional CPU implementation for same algorithm.
Our implementation with a software radio platform (explained in Section 6) is shown to achieve very low physical error rate.
We tested our implementation with two data sets from http://kdd.ics.uci.edu/databases/: Vicon Physical Action Data Set (VPA) and EEG.
In our implementation with trapped atomic ions, an auxiliary "logic" ion provides sympathetic laser cooling, state initialization, and detection for a simultaneously trapped "spectroscopy" ion.
Moreover, we also compare our implementation with the current state-of-the-art of linear SVM classifiers LIBLINEAR to validate our approach.
We cannot compare the reliability of our implementation with that of Parnin et al. because they did not present an assessment for it.
Finally, we validate our implementation with an industrial test case on 8 heterogeneous nodes, each composed of 12 CPUs and 3 GPUs.
This work also presents computational tests using benchmark instances to compare our implementation with the most competitive methods from the literature.
Even our implementation with relatively older generation Xilinx Spartan3 XC3S1500-FPGA0-4 FPGA with its static power consumption of 41 mW provides comparable power consumption performance to that of commonly used processor-based implementations.
