Are novel, complex, and specialized neural network architectures always better for language modeling? Recent papers have shown otherwise. Language models are used to predict the next token given the preceeding tokens. Most operate at word-level or character-level. Word-level models have large vocabulary sizes (how many words are there in the English language?) compared to character-level models (there are 26 letters in the English language). This means that character-level models require less memory. On the other hand, when processing a sentence, character-level models see a large number of tokens (each character is a token) compared to word-level models. A large number of tokens (long sequence) is harder for neural networks because of the vanishing gradients problem.

One of the most alarming cultural developments in the internet age has been due to the way that we access news content. The creation of filter bubbles, where people only consume media that expresses views that they are likely to agree with, has led to entrenchment of existing biases in society.

Yesterday I got to attend Man AHL’s first London Python Data hackathon (21-22 April – photos online). I went with the goal of publishing my ipython_memory_usage tool from GitHub to PyPI (success!), updating the docs (success!) and starting to work on the YellowBrick project (partial-success).

As announced yesterday, HIBPwned version 0.1.7 has been released to CRAN! Although the release was mainly a maintenance release building on Steph’s already great code, internal changes were made to start transforming HIBPwned into a real showcase of web API package development. Let’s summarize some interesting points:

** Thu 19 April 2018

The alternative title of this blog post is HIBPwned version 0.1.7 has been released! W00t!. Steph’s HIBPwned package utilises the HaveIBeenPwned.com API to check whether email addresses and/or user names have been present in any publicly disclosed data breach. In other words, this package potentially delivers bad news, but useful bad news!

A common popular technique for learning data science is starting a project. Here are the 3 E’s for why building a data science project is a good idea.

I presented a talk with this title at the Applied Machine Learning Conference at Tom Tom Fest in Charlottesville (which I also helped plan) last Thursday April 12, 2018.

Sell in May and go away. Is there any truth to this? Did some work on seasonalities recently and applied it to the stock market to quantify the truthfulness of this statement.