Variable naming inconsistency?

[theUltraMarine19]

Subject of the issue

In line 54, 55 of pgmpy/pgmpy/models/LinearGaussianBayesianNetwork.py, cpd.variables has been checked for, whereas, line 66 of pgmpy/pgmpy/factors/continuous/LinearGaussianCPD.py has self.variable specified and not variables. Hence, I think that this error is being raised. Highly obliged if you look into it.

Your environment

• pgmpy version : 0.1.6
• Python version : 3.4
• Operating System : Ubuntu 14.04

Steps to reproduce

from pgmpy.models import *
model = LinearGaussianBayesianNetwork([('hist_dist', 'dist_relevance'),('curr_dist', 'dist_relevance'),('monthly_exp','offer_interest'),('cat_merchant_affinity','offer_interest'),
('offer_interest','offer_hit_effect'),('dist_relevance','offer_hit_effect')])

model.add_nodes_from(['hist_dist', 'curr_dist', 'dist_relevance','monthly_exp','cat_merchant_affinity','offer_interest','offer_hit_effect'])

from pgmpy.factors.continuous import LinearGaussianCPD

hist_dist_cpd = LinearGaussianCPD('hist_dist', [200], 20)
curr_dist_cpd = LinearGaussianCPD('curr_dist', [400], 50)
dist_rel_cpd = LinearGaussianCPD('dist_relevance', [0, 0.5, 0.5], 0.1, ['hist_dist', 'curr_dist'])

monthly_exp_cpd = LinearGaussianCPD('monthly_exp', [45000], 5000)
cat_merchant_cpd = LinearGaussianCPD('cat_merchant_affinity', [0.6], 0.2)
offer_interest_cpd = LinearGaussianCPD('offer_interest', [0, 0.7, 0.3], 0.1, ['monthly_exp', 'cat_merchant_affinity'])

offer_hit_effect_cpd = LinearGaussianCPD('offer_hit_effect', [0.1, 0.3, 0.6], 0.2, ['dist_relevance', 'offer_interest'])
model.add_cpds(hist_dist_cpd, curr_dist_cpd, dist_rel_cpd, monthly_exp_cpd, cat_merchant_cpd, offer_interest_cpd, offer_hit_effect_cpd)

Expected behaviour

No error

Actual behaviour

File "linearGaussian.py", line 25, in
model.add_cpds(hist_dist_cpd, curr_dist_cpd, dist_rel_cpd, monthly_exp_cpd, cat_merchant_cpd, offer_interest_cpd, offer_hit_effect_cpd)
File "/usr/local/lib/python3.4/dist-packages/pgmpy-0.1.6-py3.4.egg/pgmpy/models/LinearGaussianBayesianNetwork.py", line 54, in add_cpds
if set(cpd.variables) - set(cpd.variables).intersection(
AttributeError: 'LinearGaussianCPD' object has no attribute 'variables'

[theUltraMarine19]

Error corrected : On line 78, 79 of LinearGaussianCPD.py, please put self.variables instead of variables.
All other things remaining the same.
If you grant permission, I can commit and push.

[ankurankan]

@Arijit-hydrated Thanks for reporting this. Sure, if you would like to fix this, please create a pull request.

[JZQ-Coder]

I meet the same bug, so this bug has not been fixed in the current version(0.1.11) ?

[ankurankan]

@JZQ-Coder No, it's not fixed it.

[theUltraMarine19]

@JZQ-Coder The fix is simple and code changes are in this PR #962. I found it and sent out a PR to fix it 3 years ago

[JZQ-Coder]

Glad to hear from you，I changed the source code as you prompted But there is another issuse,when i run the code Then,I find predict method has not been implemented by LinearGaussianBayesianNetWork. So can you tell me if pgmpy support Continuous variable Bayesian network in v 0.1.11? If it can do it,please give me a example. Looking forward to hearing from you

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人: "pgmpy/pgmpy" <notifications@github.com&gt;; 发送时间:&nbsp;2020年9月29日(星期二) 上午10:57 收件人:&nbsp;"pgmpy/pgmpy"<pgmpy@noreply.github.com&gt;; 抄送:&nbsp;"等风"<643187458@qq.com&gt;;"Mention"<mention@noreply.github.com&gt;; 主题:&nbsp;Re: [pgmpy/pgmpy] Variable naming inconsistency? (#961) @JZQ-Coder The fix is simple and code changes are in this PR #962. I found it and sent out a PR to fix it 3 years ago — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[ankurankan]

@JZQ-Coder No, continuous variables are not completely supported. Continuous variables/ LinearGaussianCPDs are not completely integrated with all parts of the package, so you will run into errors.