Estimating neural networks-based algorithm for adaptive cachereplacement

摘要

In this paper, we propose an adaptive cache replacement schemenbased on the estimating type of neural networks (NN's). The statisticalnprediction property of such NN's is used in our work to develop a neuralnnetwork based replacement policy which can effectively identify andneliminate inactive cache lines. This would provide larger free space forna cache to retain actively referenced lines. The proposed strategy may,ntherefore, yield better cache performance as compared to thenconventional schemes. Simulation results for a wide spectrum of cachenconfigurations indicate that the estimating neural network basednreplacement scheme provides significant performance advantage overnexisting policies