PyData London 2015

Over the weekend we presented at the PyData London 2015 conference held at Bloomberg. PyData brings together a community of folk with an interest in Python and data. The conference was split over three days with tutorials on Friday and general sessions over the weekend. We attended Saturday and Sunday.

On Saturday morning we gave an introductory talk on agent-based modelling and discussed our recent work modelling the 2011 London riots. We really enjoyed giving the talk, and we are happy to report that the room was full (around 50 people) with some people even sitting on the floor. Great to see such interest in ABM (and the riots)! Our slides are here.

Santable team in action

Sandtable team in action

To view our stop motion video of Thomas Schelling’s segregation model (mentioned in the talk) and understand more of the historical context surrounding the model, see a previous post.

"It's all about the confidence"

“It’s all about the confidence”


The first talk on Saturday was a keynote delivered by Helena Bengtsson from the Guardian. It was fascinating to get glimpse of the data work behind the news stories. It was also interesting to see that Microsoft Access is still used. Who knew.

This was followed by a talk on Voronoi diagrams by Tyler Reddy. Tyler introduced Voronoi diagrams and then discussed how they are applied to spheres and used in computational virology. Interestingly, Voronoi diagrams were used by British physician John Snow in the 1850s to demonstrate that a majority of those who died in the Soho cholera epidemic lived closer to the infected Broad Street pump than to any other water pumps.

After lunch we enjoyed the second keynote by artist Eric Drass (Shardcore). We enjoyed Eric’s talk last year, so we were anticipating another interesting presentation. Eric gave us a whiz-bang tour of the philosophy of mind. His central thesis is that limitations in the discourse around AI, for example: the definition of intelligence, are similar to those relating to the discussion of other minds in philosophy and psychology. Along the way he introduced us to Alex, your BFF bot — the online friend that we all crave. Yes, ‘we’, that includes you. She gives us the “love” and attention, and validation, that we need. Twitter shut down BFBot1; but as Eric says, you can’t keep a bot down. She’s back. We also met the Theresa May bot that just adds people to Twitter lists, such as: ‘Confirmed deviant’, ‘HMRC audit’, and ‘Candidate for extradition’. Thought-provoking.

What's it like to be a bot called Derek?

What’s it like to be a bot called Derek?

In the afternoon, Camilla Montonen gave a delightful talk about her analysis of the dynamics of the London underground using graph visualisations, network analysis and discrete-event simulation. A good candidate for an ABM perhaps?


On Sunday, we started with a well paced introduction to deep learning from Roelof Pieters. The talk included a good overview of the Python packages available for deep learning. He included images from Google’s Inceptionism – an attempt to lift the hood on the layers of deep neural networks. Machine psychedelia. Trippy.

This was followed by a talk from the trenches by Ian Ozvald on lessons from a decade of industrial data science. Some really good advice here. Definitely worth checking out. Ian’s write up and slides available here.

After lunch, we enjoyed hearing from pandas committer Jeff Reback on ‘Performance pandas’. Don’t use ‘apply’ on pandas DataFrames, or bother with ‘in_place’. We heard you Jeff! Some great tips. Well worth checking out if you use pandas – which you probably should do!

Next we heard a talk from Juan Luis Cano on Jupyter notebooks, in particular on reproducibility and the open Science movement. Too many (big) mistakes made by closed Science. Notebooks are spreading like a disease. A great disease.

Finally, it was great to see a talk by Dylan Barth and Stuart Coleman on Luigi – our favourite data pipeline framework. We believe Luigi should be widely adopted, so much so that we have intergrated it with the Sandtable Model Foundry.

As always, it was great to meet like-minded folk.

Thanks, again, to those who came to our talk, and to the organisers. No doubt see you next year!