Hi,

I tried to implement the sparse filtering in R but got this warning message, do you know what it is?
Warning messages:
1: In optimx.run(par, optcfg$ufn, optcfg$ugr, optcfg$uhess, lower, :
Hessian is reported non-symmetric with asymmetry ratio 0.551328252008331
2: Hessian forced symmetric

I believe there is a slight mistake on the demo script (sparseFilteringDemo.py). In the original MATLAB code by Ngiam, the DC is subtracted from the data as:

%% Remove DC
data = bsxfun(@minus, data, mean(data));

This subtracts the column-wise mean from the data. However, in the sparseFilteringDemo.py (line 72), the DC is subtracted as:

data -= data.mean()

This subtracts the mean of the flattened array (i.e. a scalar), not the column-wise mean. The correct code should be:

data -= data.mean(axis=0)

Same with data1 on line 82.

Thank you,

Shigeru