\cite{Dickey_2009} criterion to assess single-unit stability by measuring the similarity of average spike
waveforms and interspike interval histograms
Classifier using cross-correlograms, autocotrelogram, waveform and mean firing rate \cite{Fraser_2012} . It requires that the neurons should have some functional connectivity to display a rich cross-correlogram. In a similar way interspike interval histograms (ISIHs) have been uses as a feature, but this method should not be used when the neurons present a permanent change in firing properties \cite{Dickey_2009,Eleryan_2014}
[mas copados]
\cite{Bar_Hillel_2006} requires large samples of spikes with gaussian distribution
\cite{Wolf_2009} Looks like can't handle sparse neurons. too few spikes per interval
\cite{Shalchyan_2014} keep 100 spikes from before, windows of 10 s, perfect initial cluster, it doenst allow new clusters, have issues with merging of clusters
\cite{Shan_2017} mixture of t-distribution using windows of 1 minute. Has limitation with overclustering. Fixed number of clusters