Gaussian Auto Embeddings

Sort of came up with an alternative to VI here.

Sensors are Servers

The only servers in my stack are the sensors themselves.

Wine Cellar Strategies

Pretending there's an optimal way to drink it.

VI Drives me NUTS

Feel free to be a bit weary.

Rephrasing the Billboard

A probability problem involving 40,000 sent letters.

Amdahl's Law

The more CPU's you add, the worse it gets.

Feed Forward Posteriors

Combine the Neural with the Normal.

Passive Agressive Algorithms

A VeryGood[tm] name for a VeryGood[tm] Algorithm.

Vary Very Optimally

How to search in search space.

Bayesian Shaving Cream

A plausible (and general) method for model selection.

Bayesian/Streaming Algorithms

A convenient overlap.

Avoiding, and Preventing, Joins

Join me in preventing this.

Hello DeepQ

Stick a Network in the Q-learning algorithm.

Twain Learning

Never let your school get into the way of your regression.

Switching to Sampling in Order to Switch

A simple introduction to PyMC3.

Pokemon Recommendations, Part 2

Even more of a Sequel than SQL.

Lego Minifigures

An investment opportunity and sampling.

Custom Predictors in R

It's different than Python, but S3 isn't *that* bad.

Ensemble Yourself

Scribbles as an Algorithm Service.

Pokemon Recommendations, Part 1

An Attempt at an OptimalPortfolio[tm]/

Linear Models Solving Non-Linear Problems

XOR turns out to be a bad argument.

Variable Selection in Machine Learning

Merely *a* argument, but one that I like.

Digital Nomad

Some observations but also downsides.

Vanity Metrics

How I got an A+ for measuring the wrong thing.

More articles »