Skip to main content
. 2013 Oct 2;111(1):217–227. doi: 10.1152/jn.01046.2012

Table 3.

Optimal bandwidths for decoder variants in Table 2

Spike Feature Decoder
Identity Decoder
All spikes
Spikes in isolated clusters
Spikes in hash
MUA Decoder, All Spikes
Isolated clusters + hash
Isolated clusters
Hash only
All clusters
Data Set ha hx ha hx ha hx hx hx hx hx hx
R1D2 20 3.75 17.5 6.25 20 7.5 5 2.5 7.5 3.75 2.5
R2D2 22.5 3.75 17.5 6.25 17.5 6.25 5 2.5 6.25 6.25 5
FK11 20 6.25 25 7.5 25 8.75 17.5 5 6.25 11.25 6.25
R1D1 17.5 6.25 20 7.5 20 6.25 2.5 5 3.75 2.5 3.75
R2D1 27.5 5 20 7.5 17.5 6.25 2.5 2.5 3.75 2.5 5
Esm02 22.5 6.25 32.5 8.75 25 5 5 3.75 5 5 5
Sat2 20 8.75 20 7.5 30 7.5 8.75 8.75 6.25 7.5 8.75
SL13 20 3.75 17.5 7.5 22.5 6.25 5 3.75 3.75 12.5 8.75
SL16 30 8.75 30 7.5 20 6.25 12.5 7.5 7.5 10 6.25
Esm01 30 7.5 45 10 27.5 7.5 5 7.5 6.25 6.25 7.5
SL14 27.5 6.25 27.5 6.25 17.5 7.5 12.5 6.25 3.75 10 5
SL15 17.5 7.5 27.5 5 17.5 7.5 17.5 2.5 3.75 11.25 3.75

Data sets are ranked by number of clusters. ha and hx, Bandwidths selected for feature dimensions (spike amplitude in microvolts) and stimulus dimensions (position in centimeters), respectively.