MATLAB for scientific programming

The universe is a really, really, really vast and complex place. In fact, we understand so little of the universe that we don’t even understand how much we don’t understand. On the other hand, engineers, physicists, and biologists have developed incredible machines to measure the things inside the universe, everything from radars to thermometers to MRIs to tiny electrodes that measure the electrical impulse from individual brain cells.

How can we make sense of the increasingly big data that these machines are giving us? I bet you can guess what I’m going to write next — put the data into MATLAB and do lots and lots of analyses, modeling, and visualization!

Computer code that is optimized for data analysis, modeling, simulations, and visualizations is generally called “scientific programming,” and that is where MATLAB really excels. And this brings me back to the first paragraph of this blog: Should you learn MATLAB? If you are interested in learning or having a career in science, mathematics (abstract or applied), engineering, computational biology, physics, or data-oriented finance, then the answer is Yes, you should learn MATLAB.

I use MATLAB in two ways. First and foremost, as a way to analyze large-scale, multivariate, multidimensional datasets. I run a neuroscience (brain science) research lab, in which we measure tiny electrical impulses generated by thousands of brain cells to understand how those cells process sensory information and coordinate with each other to produce brain activity that spans multiple orders of magnitude in time and in space. And to answer your questions: Yes, brain science is hard, and yes, it’s absolutely fascinating! There is no better program than MATLAB for our work.

The second way I use MATLAB is as a tool in teaching about data analysis and mathematics. Most people do not find staring at equations to be a good way to learn advanced mathematical concepts. Instead, a better way is to translate equations into MATLAB and then simulate and visualize them. I often say “you can learn a lot of math with a little bit of coding.”

Engineers and scientists use MATLAB to organize, clean, and analyze complex data sets from diverse fields such as climatology, predictive maintenance, medical research, and finance. MATLAB provides:

  • Datatypes and preprocessing capabilities designed for engineering and scientific data
  • Interactive and highly customizable data visualizations
  • Thousands of prebuilt functions for statistical analysis, machine learning, and signal processing
  • Extensive and professionally written documentation
  • Accelerated performance with simple code changes and additional hardware
  • Expanded analysis to big data without big code changes
  • Automatic packaging of analysis into freely distributable software components or embeddable source code without manually recoding algorithms
  • Sharable reports automatically generated from your analysis

Organize and Explore Data

Organize your data with datatypes designed for tabular, time-series, categorical, and text data. Use the MATLAB language to write programs based on thousands of algorithms from a wide variety of domains. Interactively customize visualizations, then automatically generate the MATLAB code to reproduce them with new data.

Analyze Data with Less Code

MATLAB apps allow you to interactively perform iterative tasks such as training machine learning models or labeling data. These apps then generate the MATLAB code needed to programmatically reproduce the work you did interactively.

Use a prebuilt family of functions for identifying and cleaning sensor drift, signal outliers, missing data, and noise. Combine separate data sets by joining tables and synchronizing time-series data. Live Editor Tasks let you interactively solve these problems within your live script and generate the code for you. Learn how to use new MATLAB functions with extensive and professionally written documentation.

Expand Your Analysis with Few Changes

Use parfor loops and multiprocessor hardware to accelerate parallel analysis with almost no code changes. Create gpuarrays to take advantage of GPU acceleration for appropriate algorithms. Process out-of-memory data sets using tall arrays, which overload hundreds of functions throughout the data analysis workflow to operate on out-of-memory data.

Share Your Results Seamlessly

Package your analysis in freely shareable software components such as executables, C/C++ libraries, .NET assemblies, Java libraries, and Python packages. Automatically translate your MATLAB code to C and C++ code for deployment to embedded targets. Document your work using MATLAB Live Editor and export your results to reports in PDF, Microsoft Word, Latex, and HTML.

0
Would love your thoughts, please comment.x
()
x