There is much ado about variable selection or variable utility valuation in supervised machine learning. In this note we will try to disarm some possibly common fallacies, and to set reasonable expectations about how variable valuation can work. Introduction In general variable valuation is estimating the utility that a column […]

Estimated reading time: 14 minutes

Introduction Here is a quick data-scientist / data-analyst question: what is the overall trend or shape in the following noisy data? For our specific example: How do we relate value as a noisy function (or relation) of m? This example arose in producing our tutorial “The Nature of Overfitting”. One […]

Estimated reading time: 12 minutes

Introduction I would like to talk about the nature of supervised machine learning and overfitting. One of the cornerstones of our data science intensives is giving the participants the experiences of a data scientist in a safe controlled environment. We hope by working examples they can quickly get to the […]

Estimated reading time: 33 minutes

Why a mere accurate classification rule may not meet your business needs. And why you should insist on a model that returns numeric scores for classification problems. (link)

Estimated reading time: 20 seconds

Let’s please stop saying somebody isn’t a data scientist if they haven’t memorized the innards of one obscure machine learning algorithm, or blow the right smoke during an interoo (“Kangaroo interview”, thanks Jim Ruppert for this term!). Let us, instead, think of the data scientist as the bus driver. It […]

Estimated reading time: 1 minute

I am sharing some rough notes (in R and Python) here on how while dot(a, b) fulfills “Mercer’s condition” (by definition!, and I’ll just informally call these beasts a “Mercer Kernel”), the seemingly harmless variations abs(dot(a, b)) relu(dot(a, b)) are not Mercer Kernels (relu(x) = max(0, x) = (abs(x) + […]

Estimated reading time: 2 minutes

I am sharing a new free video where I work through a great common argument that bounds expected excess generalization error as a ratio of model complexity (in rows) over training set size (again in rows), independent of problem dimension. (link) For more of my notes on support vector machines […]

Estimated reading time: 34 seconds