Marius Pachitariu

It would be great if @jaib1 could redo his comparisons after we port the modified Cuda kernels from @jenniferColonell, which make the algorithm deterministic.

Due to recent developments in the algorithm, I don't think that will be necessary anymore.

I don't know, the new drift correction strategy doesn't use batch re-ordering any more or any kind of tracking, but I've only tested it on dense Neuropixels probes. I don't...

Fresh run from inside the GUI. @shashwatsridhar has also gotten this before on a different dataset, though not on this one.

There is already a temporal smoothing step in the drift correction, which smooths the "Z" cross-correlation curves before taking the max (which is the drift estimate). This effectively should work...

@Selmaan that is a very stable drift trace, and the apparent "noise" is just due to hitting the integer floor of the estimation process which is about half a micron...

Does that correspond to batches with no spikes, or very few spikes? Do you wait for the probe to settle at the beginning of the recording?

Sure, here's two that I know of: https://github.com/EnnyvanBeest/UnitMatch https://github.com/janelia-TDHarrisLab/Yuan-Neuron_Tracking

hi @Selmaan, I think this is similar to what happened for the torus paper and Richard had a modification to the merging and splitting steps that deals with it, at...

This is the last merging step [here](https://github.com/MouseLand/Kilosort/blob/1d12619093874b4ae87dfaa48f0e27d31773c96a/kilosort/run_kilosort.py#L522). It calls this function [here](https://github.com/MouseLand/Kilosort/blob/1d12619093874b4ae87dfaa48f0e27d31773c96a/kilosort/template_matching.py#L183) I have not made it this far with adding comments, so it might not be very clear yet...