Abstract Details
| Name: Suryarao Bethapudi Affiliation: Indian Institute of Technology Hyderabad Conference ID: ASI2018_1568 Title : Separation of pulsar signals from noise using supervised machine learning algorithms Authors and Co-Authors : Shantanu Desai Department of Physics Indian Institute of Technology Hyderabad Abstract Type : Poster Abstract Category : Instrumentation and Techniques Abstract : We evaluate the performance of four different machine learning (ML) algorithms: (an Ar- tificial Neural Network Multi-Layer Perceptron (ANN MLP), Adaboost, Gradient Boosting Classifier (GBC), XGBoost, for the separation of pulsars from radio frequency interference (RFI) and other sources of noise, using a dataset obtained from the post-processing of a pul- sar search pipeline. This dataset was previously used for cross-validation of the SPINN-based machine learning engine, used for the reprocessing of HTRU-S survey data. We have used Synthetic Minority Over-sampling Technique (SMOTE) to deal with high class imbalance in the dataset. We report a variety of quality metrics from all four of these algorithms on both the non-SMOTE and SMOTE datasets. For all the above machine learning (ML) methods, we report an accuracy of near 100% for both the non-SMOTE and SMOTE cases. For recall of 100%, the ANN (MLP) reports false positive rates (FPRs) of 7.59e − 4, 6.38e − 4, Ad- aboost FPRs are 2.74e − 2, 4.49e − 2, GBC FPRs are 1.63e − 4, 2.04e − 4, XGBoost FPRs are 4.55e − 4, 8.98e − 4 for the non-SMOTE and SMOTE datasets respectively. We study feature importances using Adaboost, GBC, and XGBoost and also from the minimum Re- dundancy Maximum Relevance approach to report algorithm-agnostic feature ranking. From these methods, we find that the signal to noise of the folded profile |