Since Andrej Karpathy conviced me of the The Unreasonable Effectiveness of Recurrent Neural Networks, I decided to give it a try as soon as possible.

Predicting fantasy points is hard. On average the models are only off by about 6.5 points, but can miss pretty big when players do very well or poorly, which is really what you usually care about. This is inheritely a challenging problem - how do you use past data to try and determine a future performance that differs from that past (like a break out game or a slump)? On some level, I imagine this is impossible, but think we could maybe do better. For instance, my model doesn’t account for injury at all. Knowing that a player is coming off an injury could be very valuable. Or knowing that a key offensive lineman is out could really affect a RBs performance. I think the NFL has a lot of cool work that could be done with the right data and some deep thought about how these data could be used to understand performance.

## Yet Another PhD to Data Science Post (Part III)

(I’m using AWS as an example, but this post applies just as much to other cloud providers)

This morning I helped my mum recover a lost Word document which I had reformatted for her over the weekend. Just using Word feels excruciating after writing 70,000 words across the summer in markdown.

I recently finished building a web app that recommends travel destinations. You input a country, and it provides you with 5 other countries which you might also enjoy. The recommendations are generated via a basic application of collaborative filtering. In effect, you query for a country, and the engine suggests additional countries enjoyed by other users of similar taste. The methodology is pretty simple:

Since the original Artistic style transfer and the subsequent Torch implementation of the algorithm by Justin Johnson were released I’ve been playing with various ways to use the algorithm in other ways. Here’s a quick dump of the results of my experiments. None of these are particularily rigorous and I think there’s plenty of room for improvement.

In this blog we will engineer a new feature to go into our model. So far we have predominantly been using localised features – information about pixels that are located nearby the pixel whose brightness we are predicting. In this blog we will consider the structure of a document, and use that to improve our model.

Julia is a relatively new programming language with the declared goal to become the leading language for scientific computing.