open-connectome-classes / StatConn-Spring-2015-Info

introductory material
18 stars 4 forks source link

Necessity for Spike Sorting #198

Open yaxigeigei opened 9 years ago

yaxigeigei commented 9 years ago

During the last presentation, Joshua raised the point, as I interpret, that if the dimensionality of unit(putative neuron) activity will be reduced (e.g. via PCA or ICA) anyway, why bother to sort them out in the first place? In recording, we can see a lot of non-specific background noises but there are also noises which has specific sources such as power line, stimulus artifacts, animal movements, and so on. Usually these noises have specific features and can be sorted out. In addition, even some non-specific noises can be reduced with conservative empirical criteria. Without removing them from the raw recording, the following analysis can be affected.

ajulian3 commented 9 years ago

They should be sorted out because that removes the possibility of having bias or unaccounted information from your analysis. Neuron activity is so acute so I think it would make sense to take into account this information. However, I understand your argument for sorting this out later in the process.