Towards Energy-Efficient Computing with a Precision-Scalable SAE-Subspace KNN Approach

Energy efficiency is important for all kinds of computer systems, from handhelds to supercomputers. For energy-efficient computing, it's important to have models that show how resource use and power use are related, as well as reliable measurements to evaluate system designs and find solutions that work. According to the results of this study, the right order for preprocessing, feature selection, and model training is as follows. Basic VAD, FFT, Pre-Emphasis, and framing are some of the processes that need to be taken to prepare the data for analysis. Then, to get the best features while keeping processing costs low, a dynamic recursive feature selection technique is used to get the best features. The new SAE-kNN architecture uses both SAE and subspace learning to train models. The results of the studies show that this hybrid model can classify things with 96.72% accuracy, which is better than both kNN and SAE-based techniques. These results show that the method could make things work better and faster. This method could lead to the creation of smart systems that effectively control both processing power and energy use.