Hi, I’m Alex Bird, and this blog. Welcome, and thank you for being here!

While I should probably refrain from using the first person pronoun, I have a very strong preference for speaking in person, and old habits die hard. Just for a moment, I would prefer to pretend that we’re not meeting through the emotionless medium of a screen, despite the fact that we are, and the fact that we are not really meeting. To add insult to injury, I cannot really welcome you, since I’m not doing any of the hosting. I must thank Github for that honour.1 My apologies for this rather uncongenial experience.


Who am I?

I’m just finishing up a phd in machine learning (ML) at the Turing Institute (London, UK) and the University of Edinburgh. I came into ML through a moderately circuitous path. I spent my (maths) undergraduate more interested in music than mathematics, which continued for a number of years after. Finally capitulating to a ‘real job’, I worked for 5 years in the data analytics space in credit risk and retail (riding the swell of the data science wave at that time). Modelling became my new music, and I spent all my free time in my mid/late twenties teaching myself statistics, modelling and how to code.

There’s a certain magic about machine learning, when applied right. I’m keen to pitch in wherever I can to help this disappointingly rare feat become easier and more common. This may take the form of more research, or more blogging. I also do a bit of teaching with Cambridge Spark. But this is fundamentally a community effort, and I’m currently working out how I can better serve the community in this endeavour. My interests are currently in time series models, deep generative models, multi-task/meta-learning, variational and Monte Carlo methods. I intend to touch on many of these over the coming months (and years!).


What is this?

This is primarily a blog for my thoughts and explanations of machine learning models. To be sure, the internet is full of blogs and medium posts about machine learning, some better than others. I’m intending to fill in some gaps from the research that I’ve been doing, rather than following in well trodden paths. Initially this will be my phd projects, and related learnings, much of which did not make it into any publication. Research probably bears some analogy to rejection sampling, especially where little theory is available for guidance (as in ML), and one learns as much through failures (perhaps more so) than successes. I’m especially interested in making some of my negative results available; while it is difficult to verify these, it may still give a few pointers to those who set foot in the same direction.


Why is there only a single blog post?

Yeah, so I started this blog at the end of March 2020, and I’ve managed to be ill for the entirety of April. I have a number of blog posts that I’d like to write – but I’m having to juggle writing up my phd, paid work, and taking it easy until I get better.


Contact

I’m not overly enamoured of social media, and I’m not really contactable there, so I’ll work out some way for those who are interested to get in touch. For the time being, you can use {first-letter-of-first-name} {last-name} {at} turing {dot} ac {dot} uk (no spaces), but I’ll try to get around to something better soon.


Footnotes:

  1. Also thanks to fast.ai for the site template (introductory blog). I’ve made some modifications – so any poor choices remain my own responsibility. No, I’m not using Minimal Mistakes, and yes I know it’s excellent.