Mert Turkol

Programming: C/C++, MATLAB, bash
Operating Systems: Linux, Windows
Specialties: Time Series Analysis, Digital Signal Processing, Bayesian Inference, Quantitative Research
Highlights:
• Wrote approximately 20,000 lines of compilable code (~12,250 for real-world/experimental data, ~7,750 for synthetic data)
• The largest data set worked with: 11TB images of 12 years of Astrophysical object observations
• Automated mass-processing of data from 2500+ experiments by developing code and data-handling conventions for the research group in MRELab at UofM
• Provided patent landscape analysis and technical assessment of over 150 new University inventions in the fields of numerical algorithms, engineering analysis software, and physical sciences

View my LinkedIn profile
View my GitHub profile
Listen to my music on SoundCloud

View My GitHub Profile

Portfolio


Select Projects

Please click on a project’s title above its thumbnail for a detailed report/demonstration.

Data-Adaptive, Time-Local Methodology for Numerical Modeling of Nonlinear Nonstationary Dynamical Systems

• Developed a bootstrapping procedure for Bayesian model selection in design of bandwidth-adaptive FIR filters to predict the unknown external input of dynamical systems from their response time-series

• As an application to fluid-dynamic systems, utilized time-frequency analysis and trend filtering on the estimated external forces acting on cylinders in flow-induced oscillations

• Extracted nonlinear, nonstationary patterns of driving processes pertinent to higher harmonics of fluid-structure interaction that are excluded in conventional linear theory


Structure Discovery And Pattern Recognition Using Gaussian Process Regression

• Developed Gaussian Process Regression algorithms in MATLAB for automated structure discovery in time-series through compositional kernel search

• The search procedure automatically recognizes a space of composite kernel structures which captures underlying patterns in data and enables long-range extrapolation, while optimally deciding proper model complexity using Bayesian Information Criterion

• Fully functional with Gaussian Processes for Machine Learning Toolbox, currently under development for pattern recognition in financial time-series


Data Handling Tutorial in MATLAB

35-minute tutorial ( video and code report ) on Data Handling Basics in MATLAB for ENG101: INTRODUCTION TO COMPUTERS AND PROGRAMMING class offered at the University of Michigan. The following topics are demonstrated:

During the demonstration of the topics above, core programming concepts covered are: Array Manipulation, Random Number Generation, Linear & Logical Indexing, Code Vectorization.


My Code Repositories

Please click on a title above the blockquotes to access the related GitHub repository