Classification of Parkinson’s Disease with Big Data Analytics and Ensemble Machine learning* Build a big data analytics paradigm including data exploratory analysis, data wrangling and feature construction with spark/scala/sql, hive and pig.* Use ensemble approach including both linear and nonlinear machine learning models to predict subjects with early sign of Parkinson’s Disease and patients with diagnosed PD.* Perform feature importance analysis and identify predominant clinical features.Unsupervised learning and dimensionality reduction* Two clustering algorithms, k-means and expectation maximization, and four dimensionality reduction algorithms, principal component anaylsis (PCA), independent component analysis (ICA), randomized projections, and random forests, are compared on two data sets.* R libraries were used to apply these unsupervised learning and dimensionality reduction algorithms.Assessing supervised learning algorithms* Evaluate five algorithms: decision trees, neural networks, boosting, support vector machines, and k-nearest neighbors.* R libraries were used to apply classification algorithms to two interesting data sets from the UCI machine learning repository.Predicting disposition destination in hospitalized stroke patients* Data cleaning, preprocessing, and exploratory analysis of MIMIC-III dataset.* Feature construction for time series data.* Develop machine learning models to predict likelihood of different discharge locations.Deep learning: deep Q networks (DQN)* Implement deep Q networks (DQN) using Keras and Tensorflow.* Solve the lunar lander problem in OpenAI Gym using implemented DQN algorithm.Evaluating regression model on a movie data set using R* Data preprocessing and visualization of a movie data set gathered from IMDb.* Use linear regression to predict the movie profit based on various variables (both numeric and categorical).

compressed sensing, density functional theory

Accelerating materials property predictions using machine learning* Work in a multidisciplinary university research initiative (MURI) project. Extensive experience in team-working and collaborate with different universities and groups.* Develop a similarity-based machine learning approach that can help predict the properties of a new polymer.* Develop a genetic algorithm based method for feature selection.