Keeping up to date with machine learning at Ocado
Data science and machine learning are both incredibly fast-moving fields. Every year seems to bring new breakthroughs in algorithms as well as new tools to use them.
There were over 2,000 machine learning papers submitted to Arxiv in a single month, which makes keeping up to date a daunting task, particularly in areas in which you haven’t worked for a while. Fortunately Ocado Technology recognises the importance of continued learning and development, and therefore provides a supportive environment in which to do so.
Learn from the experts
By far the most beneficial learning experience for me was the opportunity to attend the NIPS conference alongside most of the Ocado Technology Data Science team. It was a very intense seven days of talks, workshops and tutorials, and I came out with a good overview of the latest machine learning techniques, feeling inspired and with a massive pile of papers to read. Other members of Data Science also attended ICML and ICLR and brought back their knowledge and experiences to share with the rest of the team.
Being situated within easy reach of central London provides a great opportunity to regularly attend a variety of machine learning meetups, in particular the London Machine Learning Meetup and PyData are very popular amongst data science.
Learn from other teams
Ocado employs a wide range of very talented people across multiple fields, so I find it is always worthwhile to keep up to date with what others are learning. This can be difficult to do with teams you don’t interact with regularly, but Ocado has several different ways to connect teams:
- The Ocado Technology Research Conference: An internal conference where all of Ocado Technology’s research teams (Data Science, 10x, Robotics and Simulation) present what they have been working on recently. This proved not only to be an interesting learning experience, but also a great networking event to allow further collaboration and knowledge sharing.
- The Gaussian Processes reading group: A collection of Ocado employees jointly working our way through the chapters and exercises of the Gaussian Processes for Machine Learning Textbook, which is available free online.
- Slack: Enables different teams to share interesting papers and blog posts.
Learn from your team mates
There is always a natural sharing of knowledge amongst your team mates during collaboration on projects. However, not everyone can work on the same project, so it’s important to have other avenues of sharing knowledge:
- Show-and-tells. Members of the Data Science team will give a presentation / tutorial on their project or something they’ve found interesting. Past talks have covered topics such as Bayesian statistics, TensorFlow and Monte Carlo Markov Chains.
- Team hack days are a fun way to promote knowledge sharing and can teach us valuable lessons about our team.
- Lab days give us the freedom to explore topics and technologies we are interested in, and can lead to interesting projects such as our new recommendation system.
The above methods are really great ways to learn, however it is normally a good idea to spend some time going over content by yourself to solidify your understanding, particularly for any mathematical content. There are many great machine learning resources available for free online, such as Coursera or Arxiv, as well as other resources accessible through Ocado, including the large number of books available from Ocado’s libraries or via Safari.
I’ve only been working at Ocado Technology for nine months but I’ve managed to learn a lot in my time here and you could too.
Tom Adlington, data scientist