[FieldTrip] very large weight changes when running ICA

[Aitor Egurtzegi]

Dear Fieldtrip community,

I am running an ICA with the default parameters, only specifying 
cfg.method = 'runica', and I get the following, extremely large values 
for weight changes when running it.

Is there a way to fix this?

step 1 - lrate 0.000089, wchange 339727227227851.18750000, angledelta  
0.0 deg
Lowering learning rate to 6.38132e-05 and starting again.
step 1 - lrate 0.000064, wchange 6691091255065.47558594, angledelta  0.0 deg
Lowering learning rate to 4.59455e-05 and starting again.
step 1 - lrate 0.000046, wchange 57360369209.21232605, angledelta 0.0 deg
step 2 - lrate 0.000037, wchange 42285711168995.46093750, angledelta  
0.0 deg
Lowering learning rate to 2.64646e-05 and starting again.
step 1 - lrate 0.000026, wchange 1597187.26777422, angledelta  0.0 deg
step 2 - lrate 0.000026, wchange 1837915373349.00170898, angledelta  0.0 deg
step 3 - lrate 0.000021, wchange 6356182039233.50097656, angledelta 89.8 deg
step 4 - lrate 0.000015, wchange 327277436065.93859863, angledelta 44.6 deg
step 5 - lrate 0.000012, wchange 256497271977.77600098, angledelta 86.4 deg
step 6 - lrate 0.000009, wchange 94836930214915.39062500, angledelta 
81.8 deg
step 7 - lrate 0.000006, wchange 9813461305830484.00000000, angledelta 
52.6 deg
step 8 - lrate 0.000005, wchange 270943944736795648.00000000, angledelta 
82.3 deg
step 9 - lrate 0.000004, wchange 195304431631393856.00000000, angledelta 
59.3 deg

Thanks in advance,
Aitor