In neuroscience experiments it is now typical to record from hundreds of neurons in parallel from the behaving animal. However, the activity of the neurons is often highly correlated and the number of encoded cognitive variables is believed to be much lower than the number of recorded cells. In this session we will discuss the techniques used to learn the mapping between the observed neuronal activities and the underlying cognitive variables.
 
Watch the presentation here.
 
The session will have three parts:
 
  • First, I will give a brief overview of the mathematical techniques used for dimensionality reduction in neuroscience.
  • Second I will describe a few case studies where such techniques were successfully applied to understand computation in neuronal circuits.
  • Finally I will summarize our vision and recent works in this direction.
 
For the interested participants I recommend reading the review, the title of which I have borrowed for the session:
*John P Cunningham & Byron M Yu: Dimensionality reduction for large-scale neural recordings. Nat Neurosci 17(11):1500-9 (2014)
 

Join us on Zoom!