Extract deeper insights with ROC threshold visualization — All source in this article is available on Github! Feel free to reuse graphics in any context without attribution (no rights reserved). Motivation Receiver operating characteristic (ROC) curves are invaluable for understanding the behavior of a model. There is more information packed into those curves than you typically see, though. By…

It takes balls to win a war. It takes balls of steel. — I hear you. “Ball bearings aren’t that interesting,” you say. “You wrote an entire article to justify that subtitle,” you say. You caught me on the latter. As to the former, though, you’re wrong. Background Most machines move. Most machines are made of metal. This presents some logistical challenges. Just ask…

No-fuss object-oriented programming in Python Motivation In object-oriented programming (OOP), classes describe behavior (i.e., operations associated with an object) and data (i.e., attributes associated with an object). Often, though, we find that our objects are light on behavior and heavy on data. Our class might store lots of complicated information about…

An easy bug and an elegant solution to Keras Applications preprocessing — Motivation Keras Applications is an excellent source of your standard image classification architectures: ResNets, VGGs, EfficientNets, etc. But… it is pretty easy to use them incorrectly if you aren’t careful! Let’s take a peek at the fine print in the documentation. From the documentation on EfficientNet: Note: each Keras Application expects…

Clear and concise treatment of standard and custom units — Often, I found myself writing custom unit definitions in long, complex implementations. Luckily, I discovered an alternative with the library Pint! Motivation Frequently, data science projects require mixed units. Pint gives us a nice clean interface for defining and converting between units. Example use First, we need to define a unit registry. This…

An intuitive basis for understanding all things “eigen” — Motivation We often want to transform our data to reduce the number of features while preserving as much variance (i.e., the differences among our samples) as we can. …

Detailed execution logs using Python’s trace module — Motivation Breakpoints are a wonderful thing! Sometimes, though, pdb can be a hassle, and you might not have an IDE handy. Python’s trace module is here to save the day. Python’s trace allows us to log lines of source as they are executed. Setup To demonstrate the use of the module, let’s…

An introduction to wavelets and the mathematics of convolution Motivation We’ve previously discussed how to decompose a signal of interest into frequencies components using the Fourier transform. What if we want to characterize where in a signal given frequencies occur, though? …

Learn to write temporary files and folders in Python with the tempfile module. — Motivation Often, we want to store data in a file or folder briefly while we perform some operation, but we don’t necessarily want the structures to exist beyond program execution. For instance… Testing: We may want to verify that files or folders are correctly written by our software during unit tests…