Python Debugging: How to use pdb.pm()
Python’s debugger, pdb, is a great tool and one of my favorite uses of pdb is pdb.pm(). This post shows how to use pdb.pm() by working through an example of a buggy implementation of mergesort.
Cluster a Correlation Matrix (in python)
Below is a function to rearrange variables in a correlation matrix (either pandas.DataFrame or numpy.ndarray) to group highly correlated variables near each other.
It turns a correlation matrix that looks like:
Into one...
Effect size, Sample size, Significance and Power in Hypothesis Testing
When doing hypothesis testing, there are four related concepts one should keep in mind: effect size, sample size, significance threshold and power. Most people aware of hypothesis testing are likely aware of sample size and the significance threshold (typically denoted...
Scikit-learn Tree Design
Scikit-learn has a well designed, extensible implementation of decision trees. The design pays dividends for techniques that rely on decision trees such as random forest. Overall, the design is a great example of thoughtfully designed machine learning code and is...
Lazily Importing Python Modules
Recently, I found myself wanting to defer importing a python module so that it wasn’t actually imported until it was explicitly used.
After quite a bit of searching I discovered TensorFlow has a utility to lazily import modules. They...
Understanding Stochastic Gradient Descent
Until recently, I thought of stochastic gradient descent as gradient descent… with a little bit of randomness. However, when I first looked up the details of stochastic gradient descent I was confused how it could possibly work. This post is...
Narrow language features; the case for avoiding reduce
A narrow language feature leaves little ambiguity to the intent of the code. Therefore, an experienced reader can selectively skip details of the expression without loss of understanding. In contrast, a reader is forced to carefully inspect an expression that...
Exploring backoff
Back off is the amount of time to wait before retrying a failed action. There are many choices for how to back off, including constant backoff (wait 10 seconds, then try again), random backoff (wait x seconds, where x is...
Error Handling with Clojure Async
If you are reading this, there is a good chance an Exception was thrown in one of your async blocks and something unexpected happened. If you are using at least Clojure 1.6.0, your exception printed to stderr and the go...