Reading List 1: Week 52

August 27, 2018
Reading List Quant Finance Data Science

With this post I’d like to start on a (bi-) weekly reading list to recommend interesting things I stumbled upon and to reflect on my readings and what I learned.

Life is busy, as always, so we’ll see how it goes.

Data Science

  • How to select the Right Evaluation Metric for Machine Learning Models: Part 1 Regression Metrics Post about regression metrics and their usefulness depending on the objective at hand and some details and nitpicks. Read more
  • When and How to use Weighted Least Squares (WLS) Models Short introduction to weighted least squares regression that seems suitable even for heteroscedastic data (where the variance is not constant across time). WLS weights data points with a variance and thus can handle not only heteroscedastic data but even outliers to some degree. Read more (Mathematical background)
  • Land use/Land cover classification with Deep Learning A cool application to image segmentation/classification is to find and classify land patches according to their usage or even what crops are grown on it. This post details such a task with some cool details improving the model. Read more
  • Making Music: When Simple Probabilities Outperform Deep Learning Cool post about a hand-built probabilistic model for pop music generation similar to a hidden markov model. Introduces the notion of self similarity of a song to rate its structure and compare model results. Includes the source code for the project. Read more
  • Chartmaker Directory is an overview of what types of charts can be build with each language/software, including links to implementations and examples. Good to keep handy when using a new tool for visualising data, such as Amazon QuickSight! Read more

Quant

  • How to use Monte Carlo simulation with GBM One solution to the “small data” problem in algorithmic trading is to use synthetic data. This post details this so called Monte Carlo simulation based on the geometric brownian motion process to generate such data. Read more
  • Regime-Switching & Market State Modeling Showcases a generalized framework for regime switching and market state modeling that is empirically estimated by a Markov-switching ARMA process. Something that I really should learn more about! Read more
comments powered by Disqus