This is the blog of Will Badart.

Archive Series Tags Feed

Coffee Log 27

05 Nov 2018

  • #coffee

Welcome back to the coffee log. My process has developed a pretty steady holding pattern, so there hasn’t been a lot to report lately. However, I’ve finally finished my canister of pumpkin spice coffee, and have moved on to the dark roast.

My GPG Identity

31 Oct 2018

  • #cryptography
  • #gpg
  • #web of trust

I’ve finally gotten around to properly reading up on GnuPG, using keys, and managing a trusted public identity. I’ve dabbled in the past, and this dabbling has resulted in a couple of keys with my name on them being published. The TL;DR of this post is: my real GPG public key ID is FFB2398B.

Graph Convolutional Networks

26 Oct 2018

  • #data science
  • #deep learning

Convolutional Neural Networks (CNNs) are used for learning incrementally complex abstractions over raw input data. For instance, in image recognition, an early layer might learn to identify curved lines, a middle layer may learn to flag mouths, and close to the output, a layer might learn to differentiate between a smile and a frown. (For more details, please see this fantastic blog post by Ujjwal Karn, esp. §Visualizing Convolutional Neural Networks.)

Fast DataFrame Processing with Vectorized Functions

17 Oct 2018

  • #data science
  • #pandas

You may have heard about vectorized functions in the context of numpy, and that they’re faster than regular Python functions because the iteration happens in C world.

Indices are Important

15 Oct 2018

  • #data science
  • #tutorial

I recently joined the biggest data project I’ve ever been a part of at work. It’s not necessarily huge by modern big data standards, but the main element I’m working on has over 130 million records, and takes up more than 5 Gigabytes on HDFS. The main dataset for the project is over 160 GB. This post is an attempt to document and motivate the practice of conscientiously indexing your datasets to improve preprocessing performance.

1 2 3 4 5 6 7 8 9 10 11 12 13 14