This Thursday 16th April: seminar/workshop at Crick on "Understanding the brain in theory and practice" with Caswell Barry

[Goodman, Dan]

We are hosting a series of Crick Institute-sponsored neuroscience talks at Imperial & Crick on how to get experiment and theory working together better. We think it’s an exciting time to tackle this question seriously thanks to the growth of "NeuroAI" and large scale experimental projects. Each session will have a talk around these topics and an extended networking / group discussion on the questions raised (with free food provided). We will write up a position paper based on this experience and would welcome contributions.

 

The second talk is this Thursday (16^th April) from 4.30pm at The Francis Crick Institute.

 

Speaker: Caswell Barry

Title: Place cells, grid cells, and the machines that rediscovered them

Abstract:

The hippocampus and entorhinal cortex contain spatially modulated neurons — including place, grid, and border cells — that together form a neural map of space, supporting navigation and memory. Understanding what computations these cells perform has become a productive testing ground for experiment-theory dialogue in neuroscience.

 

I will describe how predictive frameworks — including the successor representation and self-supervised deep learning models that learn to predict future sensory experience — provide a principled account of spatial coding across the hippocampal formation, explaining the activity of place cells, grid cells, and the heterogeneous firing patterns of subicular output neurons. Strikingly, these representations emerge independently in artificial neural networks, suggesting that spatial geometry reflects a fundamental computational solution to navigation.

 

A recent example from our lab illustrates what becomes possible when experiment and theory are tightly coupled: by training recurrent networks on movement statistics derived from developing rat pups — rather than idealised trajectories — we can explain how the maturation of spatial representations tracks locomotor development. This works because experimental characterisation of postnatal movement directly shaped the modelling assumptions, and the resulting model generated new predictions about the developmental timeline.

 

I will use this arc as a case study to open discussion on building better collaborations between experiment and theory in neuroscience.

 

 

The talk will be around 30m long with a focus on discussions afterwards.

 

Registration for the event is free but required:

https://www.eventbrite.co.uk/e/understanding-the-brain-in-theory-and-practice-tickets-1983856488160

 

Future events in the series:

• Imperial-X / May 14^th – Florencia Iacaruso (Crick)
• Crick / June 25^th – Megan Peters (UCL)
• Imperial-X / July 16^th – Sadra Sadeh (KCL)
• Crick / September 10^th – Dan Goodman (Imperial)

 

Hope to see you there!

 

Dan Goodman