| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657 |
- Input data size [72,1076000] = 72 channels, 1076000 frames
- Finding 72 ICA components using extended ICA.
- Kurtosis will be calculated initially every 1 blocks using 6000 data points.
- Decomposing 207 frames per ICA weight ((5184)^2 = 1076000 weights, Initial learning rate will be 0.001, block size 70.
- Learning rate will be multiplied by 0.98 whenever angledelta >= 60 deg.
- More than 32 channels: default stopping weight change 1E-7
- Training will end when wchange < 1e-07 or after 512 steps.
- Online bias adjustment will be used.
- Removing mean of each channel ...
- Final training data range: -5.78337 to 5.87512
- Computing the sphering matrix...
- Starting weights are the identity matrix ...
- Sphering the data ...
- Beginning ICA training ... first training step may be slow ...
- Lowering learning rate to 0.0009 and starting again.
- Lowering learning rate to 0.00081 and starting again.
- Lowering learning rate to 0.000729 and starting again.
- Lowering learning rate to 0.0006561 and starting again.
- Lowering learning rate to 0.00059049 and starting again.
- Lowering learning rate to 0.000531441 and starting again.
- Lowering learning rate to 0.000478297 and starting again.
- Lowering learning rate to 0.000430467 and starting again.
- Lowering learning rate to 0.00038742 and starting again.
- Lowering learning rate to 0.000348678 and starting again.
- Lowering learning rate to 0.000313811 and starting again.
- Lowering learning rate to 0.00028243 and starting again.
- Lowering learning rate to 0.000254187 and starting again.
- Lowering learning rate to 0.000228768 and starting again.
- Lowering learning rate to 0.000205891 and starting again.
- Lowering learning rate to 0.000185302 and starting again.
- Lowering learning rate to 0.000166772 and starting again.
- Lowering learning rate to 0.000150095 and starting again.
- Lowering learning rate to 0.000135085 and starting again.
- Lowering learning rate to 0.000121577 and starting again.
- Lowering learning rate to 0.000109419 and starting again.
- Lowering learning rate to 9.84771e-05 and starting again.
- Lowering learning rate to 8.86294e-05 and starting again.
- Lowering learning rate to 7.97664e-05 and starting again.
- Lowering learning rate to 7.17898e-05 and starting again.
- Lowering learning rate to 6.46108e-05 and starting again.
- Lowering learning rate to 5.81497e-05 and starting again.
- Lowering learning rate to 5.23348e-05 and starting again.
- Lowering learning rate to 4.71013e-05 and starting again.
- Lowering learning rate to 4.23912e-05 and starting again.
- Lowering learning rate to 3.8152e-05 and starting again.
- Lowering learning rate to 3.43368e-05 and starting again.
- Lowering learning rate to 3.09032e-05 and starting again.
- Lowering learning rate to 2.78128e-05 and starting again.
- Lowering learning rate to 2.50316e-05 and starting again.
- Lowering learning rate to 2.25284e-05 and starting again.
- Lowering learning rate to 2.02756e-05 and starting again.
- Lowering learning rate to 1.8248e-05 and starting again.
- Lowering learning rate to 1.64232e-05 and starting again.
- step 1 - lrate 0.000016, wchange 2185365000731481.00000000, angledelta 0.0 deg
- Lowering learning rate to 1.18247e-05 and starting again.
- step 1 - lrate 0.000012, wchange 111303566458.68484497, angledelta 0.0 deg
|
