YAO SONG

**** ***** ***** **********, ** *****

984-***-**** adkd9v@r.postjobfree.com

OBJECTIVE

An analytical and thoughtful PhD candidate in Statistics with extensive research capabilities. Excellent mathematical and computational skills. Applying for an internship in Statistics/Data Science. EDUCATION

Rutgers, The State University of New Jersey 2018 - 2023 (Anticipated) PhD in Statistics. Current GPA: 4.0/4.0, 2019 Best Performance in Ph.D Qualify Exam Duke University 2015-2017

Master of Biostatistics. GPA: 3.96/4.0, Graduated with Chair’s Academic Recognition Award Southwestern University of Finance and Economics 2011-2015 Bachelor of Science, Statistics. GPA: 4.0/5.0, Excellent Graduate TECHNICAL SKILLS

Analytical Skills Machine Learning, Predictive Model, Experimental Design, Time Series Categorical Data Analysis, Survival Analysis, Longitudinal Analysis Statistical Genetics, Generalized Linear Model

Programming R, Python, SAS, SQL, Matlab, Latex

WORK EXPERIENCE

Rutgers, The State University of New Jersey 2019 - 2020 Research Assistant

Evaluated the performance of di erent methods i.e., penalized regression, random forest, Gaussian process regression, on predicting a material data with 45 inputs and 499 outputs in R

Performed optimization methods i.e. L-BFGS-B, GenSA, to nd new input combinations with smaller loss error

Documented analytical and validation results, and gave regular presentations to the department Rutgers, The State University of New Jersey 2018 - Present Teaching Assistant

Hold weekly o ce hours to help answer students’ course-related questions and homework problems

Lead review sessions to teach 1st

-year PhD students course materials for qualify exams Duke University, Biostatistics Core 2016 - 2017

Biostatistician Intern

Cleaned and performed analysis on Observational Data and Electronic Health Record Data by using R/SAS

Developed Statistical Analysis Plans (SAP) to assist project planning and monitored the progress of projects

ACADEMIC PROJECTS

A Classi cation-Incorporated Minimum Energy Design 2019 - 2020

Built models to predict material properties total loss error in ReaxFF system with space exploration design, providing new parameter combinations close to property targets

Generated a Minimum Energy Design with a trained Deep Learning model by Python to explore desired space

Provided 50% reduction in the total loss errors in a ReaxFF systems, comparing with the conventional method

Tensegrity Robot Locomotion Simple Model Prediction 2020

Derived a space- lling design to simulate natural oscillation locomotion data

Performed Gaussian Process Regression in Python to predict 1-step lag locomotion data for each state, which improved 25% 40%in prediction mean square error, comparing with Recurrent Neural Network

Multiple-output model is in progress by considering correlations among all state positions Comparison Among Latent Factor Model and its Extensions 2019

Performed Latent Factor Models in Python to the dataset MovieLens with 100836 ratings and 3683 tag applications across 9742 movies, for building a Recommender System

Improved the model by adding regularization and bias terms

Evaluated di erent models by rating prediction evaluation i.e. RMSE and item recommendation eval- uation i.e. precision, recall or F1

Pokemon Go Prediction 2016

Performed principle componenent analysis (PCA) using R to reduce the feature dimension from 208 to 56

Utilized over-sampling and under-sampling methods to modify the imbalanced data into balanced dis- tribution

Implemented Random Forest and Boosting algorithms with Cross validation in R to nd the best prediction model with above 80% accuracy

PUBLICATION

1. Song, Y., Sengul, M., He, L.L., van Duin, A., Hung, Y., Dasgupta, T., CLAIMED: A CLAssi cation- Incorporated Minimum Energy Design to explore a multivariate response surface with feasibility constraints, arXiv preprintarXiv: 2006.05021, 2020 2. Sengul, M., Song, Y., Nayir, N., Gao, Y.W., Hung, Y., Dasgupta, T., van Duin, A. An Ini- tial Design-enhanced Deep Learning-based Optimization Framework to Parameterize Multicomponent ReaxFF Force Fields, Submitted to Journal of Chemical Theory and Computation, 2020

CERTIFICATION

SAS Certi ed Base Programmer for SAS 9 2016

SAS Certi ed Advanced Programmer for SAS 9 2016

Contact this candidate