YASS is a spike sorting pipeline developed for high-firing rate, high-collision rate retinal recordings. YASS employs a number of largely automated approaches to isolate single neuron templates and match them to the raw data. In monkey retinal recordings we found that YASS can identify dozens to hundreds of additional neurons not identified by human or other sorters and that such additional neurons have receptive fields and can be used to more accurately decode images.
The NeuroCAAS implementation of YASS works with version 2.0 of YASS.
Inputs:
Raw data: (bin file) The .bin file called 'data.bin'
Probe geometry (txt file): 'geom.txt' The ASCII txt file containing the X and Y coordinates of each electrode. The default NNs should be sufficient for most datasets that have an approximately hexagonal pattern. For novel layouts, consult https://github.com/paninski-lab/yass/wiki/Neural-Networks---Loading-and-Retraining.
config: (yaml) a yaml file containing the necessary parameters to run YASS. See https://github.com/paninski-lab/yass/wiki/Data-Format-and-Configuration-Parameters for a detailed outline
Outputs:
preprocess: contains the standardized recording and the numpy arrays used in standardization and filtering
block_1: contains results from detection and clustering, each of which includes spike_train.npy (a 2D array matching time values to assigned clusters) and templates.npy (a 3D array containing the number of units, number of time steps for each spike, and number of channels)
block_2: contains similar data to block_1
pre_final_deconv: contains time-varying templates, spike train, and soft assignments (probability that a given unit is assigned incorrectly) \n final_deconv: contains time-varying templates, spike train, and soft assignments
output: contains time-varying templates, spike train, soft assignments, and a similarity matrix indicating over-split/under-split probabilities
nn_train: contains the neural networks used for spike denoising and detection
You must login to use an analysis.