Uncovering the mechanisms behind rapid motor adaptation in the brain using recurrent neural networks

Animals and humans learn some motor skills very slowly: for example, it takes about a year for a baby to learn to walk. In comparison, motor adaptation can be very fast.

For example, the first time someone walks on snow, it might take them a little time to get used to the new conditions, but not a whole year. And adaptation depends on previous acquired skillset.

Mastering tennis for instance increases how fast squash is learned, but also some “bad habits” from tennis have to be unlearned, and switching to squash for too long can hurt one’s performance in tennis. But how the brain learns different motor skills is still unknown.

The main reasons are 1) it is very hard to record single neural activity for the time it takes to learn a new skill, and 2) bottom-up methods from computational neuroscience are very precise for modelling neurons and synapses (i.e., the connections between neurons), but fall short when it comes to learning behaviourally relevant tasks.

In contrast, methods from machine learning are becoming increasingly more powerful.My innovative idea is that instead of focusing on skill acquisition, which is slow, I propose to focus on rapid adaptation where recordings during its whole duration is feasible.

I propose to use data-driven machine learning methods alongside newly acquired motor datasets to uncover where changes occur in the brain during motor adaptation, which brain regions are involved, which rules govern the changes, and how different motor skills interact.

Unravelling the fundamental mechanisms underpinning rapid motor adaptation will equip me with the knowledge necessary to engineer tools that accelerate motor adaptation, particularly for medical applications. 100 million people in the EU alone suffer from a disability including movement disorders.

I propose to leverage these tools to accelerate how quickly users can adapt to intracortical Brain-Computer-Interfaces to improve quality of live and increase patient adoption rate.