MACHINE LEARNING, PREDCITIVE & STATISTICAL MODELING
Machine learning (ML), unsupervised and supervised; Reinforcement learning; Deep learning (DL), deep neural networks (Keras, Tensorflow, Mxnet).
Creating predictive models and data pipelines, on textual, visual, audio, and structured data.
Creating predictive models, on massive data, via distributed statistical estimation and optimization, using Hadoop and Spark (SparkR and MLLib), TensorFlow, also SAS High-Performance Procedures, achieving out-of-sample exceptional performance.
Surpassing 99% AUC-ROC for predictive models on multi-billion record datasets.
Statistical predictive modeling [20+ years], using R SAS Stata SPSS, utilizing a variety of statistical methods.
Anomaly detection (robustEM), bayesian network (bnlearn), neural network (neuralnet, deepnet), hierarchical clustering (hybridHclust), support vector machines (sparseSVM), principal components analysis (bootSVD), latent variable models (lava).
Generalized linear models (including logistic regression: binary, multinomial, ordinal), generalized estimating equations (including log-linear models), generalized additive models; longitudinal time-series modeling, survival analysis (including multiple events, competing risks, frailty, state-space and Markov chain), and repeated measures (random-effects models and marginal models), risk analysis; GIS modeling.
Extensive experience fitting wide array of quantitative models on biomedical data.