The basic approach to solving the stability problem across days was based on comparing the waveform and other metrics (i.e. mean firing rate) of the neurons detected on non-overlapped time windows and track the neurons that conserve the approximately the same proprieties \cite{Tolias_2007,Fraser_2012,Emondi_2004,Eleryan_2014,Dickey_2009}. This type of approach requires that the recording is stable enough that the waveforms are similar enough and/or that the firing dynamic of the neurons does not change. Furthermore they can not handle errors in the sorting algoritm used to get the neurons in each window.
Another family of methods uses a Bayesian framework to model blocks of spikes and then apply previous information and transition probabilities to track the neurons \cite{Bar_Hillel_2006,Wolf_2009,Shalchyan_2014,Shan_2017} . Some of these methods requires a fixed number of neurons across the recording \cite{Shalchyan_2014,Shan_2017}, and others use a mehot with high computational cost to handle a variable number of neurons \cite{Wolf_2009,Bar_Hillel_2006}.
\cite{Dhawale_2017} Sucesives SPC with all the detected spikes, NO MANEJA OVERLAPPING SPIKES