Reviews: Fast and accurate spike sorting of high-channel count probes with KiloSort

Improvements in spike sorting are always appreciated because of the always-increasing electrode array capabilities (e.g number of electrodes or density) leading to new computational challenges. Methodologically, their contribution is the observation that one can take the SVD of the mean of raw traces in the cluster order to i)regularize shape of templates, ii)obtain computational gains via a GPU implementation of the operations in equation 6. This methodological contribution leads to impressive improvements in time while maintaining comparable levels of accuracy. In general it is a very nice paper. However, I think this paper has major issues and it is not suitable for publication in the NIPS conference/proceedings. 1)Overall, it gives the impression that it hasn't been fleshed out enough.