Using Neural Network for the Evaluation of Power Consumption of Instructions Execution
Andrii Borovyi, Vasileios Konstantakos, Volodymyr Kochan, Volodymyr Turchenko, Anatoliy Sachenko, Theodore Laopoulos · 2008
In this work a method is being proposed for estimating the power consumption of digital processing systems by the use of neural networks. The case study is an ARM7TDMI processor. Real hardware data are already known for this processor and provided for neural network training. Many different attempts for training have been made, by combining different sets of training vectors to the neural network and initial results have been extracted. Results indicate that the proposed approach is good for power consumption estimation, and with a proper selection of training vectors, the neural network can provide results with increased accuracy.