log_sub-018_ica1_16-Nov-2021.txt 821 B

12345678910111213141516
  1. Input data size [72,1077500] = 72 channels, 1077500 frames
  2. Finding 72 ICA components using extended ICA.
  3. Kurtosis will be calculated initially every 1 blocks using 6000 data points.
  4. Decomposing 207 frames per ICA weight ((5184)^2 = 1077500 weights, Initial learning rate will be 0.001, block size 70.
  5. Learning rate will be multiplied by 0.98 whenever angledelta >= 60 deg.
  6. More than 32 channels: default stopping weight change 1E-7
  7. Training will end when wchange < 1e-07 or after 512 steps.
  8. Online bias adjustment will be used.
  9. Removing mean of each channel ...
  10. Final training data range: -50.9061 to 57.4365
  11. Computing the sphering matrix...
  12. Starting weights are the identity matrix ...
  13. Sphering the data ...
  14. Beginning ICA training ... first training step may be slow ...
  15. Data has rank 71. Cannot compute 72 components.