FPGA-Based Implementation of Stochastic Configuration Networks for Regression Prediction

Abstract

The implementation of neural network regression prediction based on digital circuits is one of the challenging problems in the field of machine learning and cognitive recognition, and it is also an effective way to relieve the pressure of the Internet in the era of intelligence. As a nonlinear network, the stochastic configuration network (SCN) is considered to be an effective method for regression prediction due to its good performance in learning and generalization. Therefore, in this paper, we adapt the SCN to regression analysis, and design and verify the field programmable gate array (FPGA) framework to implement SCN model for the first time. In addition, in order to improve the performance of the SCN model based on the FPGA, the implementation of the nonlinear activation function on the FPGA is optimized, which effectively improves the prediction accuracy while considering the utilization rate of hardware resources. Experimental results based on the simulation data set and the real data set prove that the proposed FPGA framework successfully implements the SCN regression prediction model, and the improved SCN model has higher accuracy and a more stable performance. Compared with the extreme learning machine (ELM), the prediction performance of the proposed SCN implementation model based on the FPGA for the simulation data set and the real data set is improved by 56.37% and 17.35%, respectively.

Efficient Neural Networks on the Edge with FPGAs by Optimizing an Adaptive Activation Function

“…Ngah et al (2016) [33] designed a differential look-up table-based sigmoid function with limited accuracy. Gao et al (2020) [34] designed a better sigmoid function on an FPGA for regression problems, but it operates at a slow speed. Xie et al (2020) [35] implemented a twofold LUT-based tanh function, but it comes at a high cost in hardware.…”

FPGA handwritten digit recognition system based on LSTM algorithm

“…Minst data set is selected to test the accuracy of each model. The Table 1 shows the comparison between this paper and CPU, GPU and literature [3]and [10] in terms of model prediction time and accuracy. Among them, in…”

An approximate randomization-based neural network with dedicated digital architecture for energy-constrained devices

“…As a result, the training procedure had to be executed for every single chip. This also held for mixed-signal realizations, which featured a dispersion of parameters, thus providing high performances [41] and supporting ultra-low-power implementations of the RBN inference step [31,42,43] Purely digital implementations for resource-constrained devices were presented in [7,8,21,44]. Those works explored design strategies that removed the need for multipliers in digital architectures.…”

“…Gao et al [44] presented an implementation of RBNs based on a three-stage architecture. The network was deployed on a Xilinx XC7 with a 50MHz clock frequency.…”