PCE with TruncNormal input variables cannot be used for uncertainty quantification and sensitivity analysis

[luyipin123]

When TruncNormal is selected as the distribution and fitting PCE, we perform chaospy.E(approx, distribution) or chaospy.Std(approx, distribution). Finally the errors are as follow:

Traceback (most recent call last):
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\distribution.py", line 618, in mom
out = [self._get_mom(kdata) for kdata in K.T]
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\distribution.py", line 618, in
out = [self._get_mom(kdata) for kdata in K.T]
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\distribution.py", line 641, in _get_mom
ret_val = float(self._mom(kdata, **parameters))
File "D:\Anaconda\lib\site-packages\chaospy\distributions\operators\joint.py", line 161, in _mom
output *= self._owners[unique_idx][1]._get_mom(kloc[index == unique_idx])
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\distribution.py", line 641, in _get_mom
ret_val = float(self._mom(kdata, **parameters))
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\shift_scale.py", line 109, in _mom
for key, coeff in zip(poly.exponents, poly.coefficients)])
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\shift_scale.py", line 109, in
for key, coeff in zip(poly.exponents, poly.coefficients)])
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\distribution.py", line 641, in _get_mom
ret_val = float(self._mom(kdata, **parameters))
File "D:\Anaconda\lib\site-packages\chaospy\distributions\baseclass\simple.py", line 110, in _mom
"%s: does not support analytical raw moments." % self)
chaospy.UnsupportedFeature: trunc_normal(lower=0, upper=1, mu=0, sigma=1): does not support analytical raw moments.

During handling of the above exception, another exception occurred:

[jonathf]

Do you mind reporting stats on your run. In particular:

• python --version
• python -c "import chaospy as cp; print(cp.__version__)"
• python -c "import numpoly as nu; print(nu.__version__)"
• python -c "import numpy as np; print(np.__version__)"

[luyipin123]

* python C3.7 * cp C4.3.3 * nu C 1.2.3 * np C1.20.3 My test_run is as follow: import chaospy as cp import numpy as np m = 2 expansion_order = 3 level = 3 def model_solver(q): return q[0]*np.e**-q[1]+1 distribution = cp.Iid(cp.TruncNormal(0, 1), m) abscissas, weights = cp.generate_quadrature(level, distribution, rule='gaussian') solves = np.array([model_solver(q) for q in abscissas.T]) expansion, norms = cp.generate_expansion(expansion_order, distribution, retall=True) approx = cp.fit_quadrature(expansion, abscissas, weights, solves, norms=norms) expected = cp.E(approx, distribution) deviation = cp.Std(approx, distribution) First_indices = cp.Sens_m(approx, distribution) Total_indices = cp.Sens_t(approx, distribution) 从 Windows 版邮件<https://go.microsoft.com/fwlink/?LinkId=550986>发送 发件人: Jonathan ***@***.***> 发送时间: 2021年8月17日 16:43 收件人: ***@***.***> 抄送: ***@***.***>; ***@***.***> 主题: Re: [jonathf/chaospy] PCE with TruncNormal input variables cannot be used for uncertainty quantification and sensitivity analysis (#354) Do you mind reporting stats on your run. In particular: * python --version * python -c "import chaospy as cp; print(cp.__version__)" * python -c "import numpoly as nu; print(nu.__version__)" * python -c "import numpy as np; print(np.__version__)" ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#354 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AVHPURHHHUVQH32D67RMN73T5IOMBANCNFSM5CJKTLNQ>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email>.

[jonathf]

I am unable to replicate the issue on my local machine.

To weed out a small list of subtle problems, do you mind doing the following.
I don't use windows (which I assume you do from the error output), so I appologies in advance for any mistake in the recipy.

Anyways:

• Make a completly new folder and copy the test_run.py script into it.
• From your shell (cmd.exe or Powershell) move into the new folder
• Make virtual env: python -m venv venv
• Make sure pip works: venv/bin/python -m ensurepip --upgrade
• Install chaospy: venv/bin/python -m pip install chaospy
• Run script: venv/bin/python test_run.py

Let me know how it goes and if it solves the problem or not.

[luyipin123]

Hello, I've followed the steps you said, but I still haven't solved the problem. I'm still prompted before the error 从 Windows 版邮件<https://go.microsoft.com/fwlink/?LinkId=550986>发送 发件人: Jonathan ***@***.***> 发送时间: 2021年8月18日 22:30 收件人: ***@***.***> 抄送: ***@***.***>; ***@***.***> 主题: Re: [jonathf/chaospy] PCE with TruncNormal input variables cannot be used for uncertainty quantification and sensitivity analysis (#354) I am unable to replicate the issue on my local machine. To weed out a small list of subtle problems, do you mind doing the following. I don't use windows (which I assume you do from the error output), so I appologies in advance for any mistake in the recipy. Anyways: * Make a completly new folder and copy the test_run.py script into it. * From your shell (cmd.exe or Powershell) move into the new folder * Make virtual env: python -m venv venv * Make sure pip works: venv/bin/python -m ensurepip --upgrade * Install chaospy: venv/bin/python -m pip install chaospy * Run script: venv/bin/python test_run.py Let me know how it goes and if it solves the problem or not. ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#354 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AVHPURFV42GWXQVEIR6VHIDT5O7YBANCNFSM5CJKTLNQ>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email>.

[jonathf]

Very weird.

Can you post the new trace stack? It shouldn't be exactly the same. The paths at least should point to a new files.