The Readback Signal Prediction based on Machine Learning Technique in Bit Patterned Media Recording System

Abstract

The Bit patterned media recording (BPMR) system is the new challenge technology for the magnetic recording systems to be produced in the future. The readback signal of the BPMR system included 2D interference as inter-track interference (ITI) and inter-symbol interference (ISI) to decrease system performance. Therefore, the data of these readback signals were interesting for finding the new model technique from machine learning to analyze about characteristics of the actual readback signal without ISI and ITI. Previous work used the machine learning technique to describe the readback signal of the Two-dimensional magnetic recording (TDMR) system. The TDMR is one of the new challenging technology for magnetic recording systems also. Therefore, in this work, we would like to use the machine learning technique such as K-Neighbors, Decision Tree, Random Forest, AdaBoost, Logistic Regression, Deep Learning, etc. A classification model is used to predict the actual readback signal output without 2D interference, track misregistration (TMR) and position jitter problems in the BPMR system. The new models can predict and perform accuracy percentage more than 90% from all models in the simulation result, especially in Deep Learning, can achieve higher accuracy than 99% and lower loss.