Optimization for Statistics

Notes on Optimization for Statistics

A large variety of statistical problems are essentially solutions to optimization problems. The mathematical techniques of optimization are fundamental to statistical theory and practice. I was fortunate to take the Optimization for Statistics special course (thanks to Dr. Ravi Varadhan and Dr. Vadim Zipunnikov) at Johns Hopkins. Here are some of my thoughts and notes on gradient descent, constrained optimization, EM Algorithms, and convex optimization, etc. )

Deep Learning for R Users

Notes on Deep Learning with R

Keras is a high level neural networks API developed by Francois Chollet to enable fast experimentation, which turns out to be extremely useful for statisticians and data scientists like me, who likes to “go from ideas to result with the least possible delays”. It has been made available in R since 2017 (see keras on CRAN). As an R programmer, I recommend starting with the two books, Deep Learning with R by Chollet and Deep Learning with R by Ghatak, since they both have details about the methodologies as well as R codes. Here are some of my notes.

Machine Learning and Statistical Modeling in tidyverse and tidymodels

Notes on Statistical Modeling and Machine Learning in R

The tidyverse is a collection of R packages designed for data science. All packages share an underlying design philosophy, grammar, and data structures. As a modern-time R programmer, there is no reason for us not to use it. I have gradually changed my coding habits and switched to tidy way of programming. Here are some of the most fundamental skills in this domain. The tidyverse is a collection of R packages designed for data science. All packages share an underlying design philosophy, grammar, and data structures. As a modern-time R programmer, there is no reason for us not to use it.

Specifically, the tidyverse and tidymodel frames make model training, prediction, interpretation more streamlined. Moreover tidymodel and caret provided model training and prediction APIs. This notes contains tutorials with R codes to fully take advantages of the tidyverse and tidymodel frameworks for statistical models and machine learning.