taku-y / bmlingam Goto Github PK

I have installed bmlingam which works great finding causality .
However I am facing the following issue while running bmlingam-coeff.

Traceback (most recent call last):
File "/home/administrator/anaconda2/bin/bmlingam-coeff", line 11, in <module>
from bmlingam.commands.bmlingam_coeff import estimate_coeff_posterior, \
ImportError: cannot import name estimate_coeff_posterior

Can you please help what must have gone wrong

Hi ,
I am facing following error while trying to run bmlingam with gauss.
Commands we have issued :

1. bmlingam-causality --standardize_on --prior_indvdls=gauss time_2.csv
2. bmlingam-causality --standardize_off --prior_indvdls=gauss time_2.csv
3. bmlingam-causality --prior_indvdls=gauss time_2.csv
4. bmlingam-causality --standardize_off --fix_mu_zero_off --max_c=1.0 --n_mc_samples=1000 --prior_scale=log_normal --L_cov_21s=[-0.9, -0.7, -0.5, -0.3, 0, 0.3, 0.5, 0.7, 0.9] --prior_indvdls=gauss time_2.csv --sampling_mode cache_mp8 --cs=[ 0. , 0.2, 0.4, 0.6, 0.8, 1. ]

In the 4th command, BMLingam does not take lists 'cs' and L_cov_21s.

Following is the stack trace.
Traceback (most recent call last): File "/usr/local/bin/bmlingam-causality", line 18, in <module> bmlingam_causality_(**params) File "/usr/local/lib/python2.7/site-packages/bmlingam/commands/bmlingam_causality.py", line 281, in bmlingam_causality varnames[list(pair)]) for pair in pairs] File "/usr/local/lib/python2.7/site-packages/bmlingam/infer_causality.py", line 24, in infer_causality find_best_model(xs, hparamss, sampling_mode) File "/usr/local/lib/python2.7/site-packages/bmlingam/find_best_model.py", line 41, in find_best_model for hparams in hparamss]) File "/usr/local/lib/python2.7/site-packages/bmlingam/bmlingam_np.py", line 56, in comp_logP cache, prior_indvdl, L_cov_21, df_indvdl, beta_coeff) File "/usr/local/lib/python2.7/site-packages/bmlingam/cache_mc.py", line 165, in fetch_mu_indvdl ('mu_indvdl1', prior_indvdl, L_cov_21, df_indvdl, beta_coeff)] KeyError: ('mu_indvdl1', 'gauss', -0.9, 8.0, None)
Thanks for looking to the problem.

Before update, need to confirm that PyMC3 is automatically installed when executing pip install BMLiNGAM-0.1.5-py2.py3-none-any.whl.

I was able to get this running with the port of PyMC3 to Aesara. I also added, to bmlingam_causality.py an exclusion of rows with any na values from pairs. It may not be appropriate for this repo but I figured I'd mention the fact since I did this upgrade due to the fact that none of my attempts to install bmlingam worked due to version conflicts involving theano.

Hi, thanks for your work.
The 'cache' version can work. But when I change the 'sampling_mode' to 'normal'. The error message 'KeyError: 'L_cov' appeared in L_cov = hparams['L_cov']. Could you please help me?

Create docker file to automate tests the library.

The CSV file I am giving has 37 variables out of which few have been processed and log has been captured, but suddenly my system crashed.
Now in my logs I can see
{
2 * log(p(M)) - log(p(M_rev)) : 22.123
Infered Causality : "A->B"
}
Is there a way to resume BMLiNGAM where the program last left with ?