Running the jupyter notebook http://nbviewer.jupyter.org/github/Effective-Quadratures/Effective-Quadratures/blob/master/ipython-notebooks/PolynomialChaosPDFs.ipynb gives a warning when plotting Parameters:

usr/local/lib/python2.7/dist-packages/equadratures-5.1-py2.7.egg/equadratures/plotting.py:168: MatplotlibDeprecationWarning: The set_axis_bgcolor function was deprecated in version 2.0. Use set_facecolor instead.
  ax.set_axis_bgcolor('whitesmoke')

It seems that the eq.datasets.load_eq_datasets() function is not able to use the correct url to download dataset for windows, it converts '\' to '%5C', due to which the function fails to work.

The output error is

400 Client Error: Bad Request for url: https://raw.githubusercontent.com/Effective-Quadratures/data-sets/main%5Cprobes%5CREADME.md

Either there is a bug here or with qr col pivoting...ipynb answers aren't quite right.

In preparation for python 3.8+ compatibility, a number of is and is not checks in poly.py and basis.py need replacing. See below...

The compiler now produces a SyntaxWarning when identity checks (is and is not) are used with certain types of literals (e.g. strings, numbers). These can often work by accident in CPython, but are not guaranteed by the language spec. The warning advises users to use equality tests (== and !=) instead. (Contributed by Serhiy Storchaka in bpo-34850.)

It seems that np.matrix will be deprecated in the future. Thus, all instances of np.matrix should be changed to np.ndarray (e.g. the output of evaluatePolyFit)

In addition, we should stipulate that all 1-d outputs be represented by 1-d arrays instead of "column vectors" with size = (length, 1). It makes working with them more intuitive.

Check this.

Current data structures used for Sobol' and related indices are awkward and unexpected.
Need to think about the workflow of using stats.

It seems that the documentation mentions plotting utilities but I can't find them in the code base, or use them with the package. They are similarly present in the Parameter documentation as well. I'm not sure whats the actual source of this documentation on the equadratures website to PR and fix this.

The quadrature rule generation seems to break when the order is increased above 4. This occurs for both tensor order and total order grids. See the attached python 2.7 code to replicate the bug. (included as a .txt file)
bug_report.txt

Spam doesn't output all the coefficients. Need to sort this out.

Currently the setup specifies a dependency of 'scipy >= 0.15.0'. However the optimisation module imports scipy.linalg.null_space which was added later (appears to have been introduced in Scipy 1.1.0: https://docs.scipy.org/doc/scipy/reference/release.1.1.0.html).

Not that it's an issue for you but our internal python environment is currently limited to scipy 0.18.1 so the fix for us isn't as straightforward as just updating the scipy package we use. On our local copy of the code I've needed to comment out the import and manually implement the null_space function. My preference for a (somewhat ugly) fix would probably be to leave the dependency as is but to wrap the import in a try/except with a local implementation of null_space defined if it doesn't work.

Also I think datasets.py imports requests which isn't currently listed as a dependency. It works with requests 2.11.1.

In the tutorial Multi-Index inside the official Documentation, instead of using the class IndexSet, Basis has to be replaced together with the use of the method 'sort' to calculate the attributes 'elements';
also, another little comment about the tutorial Polynomial Regression for time varying data: it does not report the lines in which the array yy is created (normalization of volume values).
Thanks a lot !

To improve coverage, perhaps we should remove some unused functions:

• utils.py: Everything except columnNormalize, cell2matrix, evalfunction, find_repeated_elements. evalfunction could be replaced with the numpy's apply_along_axis.
• qr.py: Everything except solveCLSQ?
• polyreg.py: F_stat (I will do this)
• polylsq.py: Merge into polyreg?

Will include more later...

Having installed with sudo python3 setup.py install importing eqaudratures fails as follows:

jshaw@swpc215-vm ~ $ ipython3
Python 3.5.3 (default, Jan 19 2017, 14:11:04) 

In [1]: from equadratures import *
---------------------------------------------------------------------------
ImportError                               Traceback (most recent call last)
<ipython-input-1-40b72a151ef6> in <module>()
----> 1 from equadratures import *

/usr/local/lib/python3.5/dist-packages/equadratures-5.1-py3.5.egg/equadratures/__init__.py in <module>()
      1 'Initialization file'
----> 2 from polyint import Polyint
      3 from parameter import Parameter
      4 from polylsq import Polylsq
      5 from polyreg import Polyreg

ImportError: No module named 'polyint'

This is using version 215df25. The same code installs with python 2.7.13 without issue.

Tentative idea: Load the Poly class:

def __init__(self, parameters, basis, task):

        # Tasks:
        # 1. uncertainty quantification --> doe
        # 2. dimension reduction --> doe
        # 3. optimization --> doe
        # 4. regression --> data 
        # 5. read in old file

As part of this version, it would be useful to create a file format that can read-in a prior polynomial.

In the implementation given below, the polynomial class needs to know the attributes of other classes--i.e., basis and parameter. It may be better if each of the other classes is endowed with lxml functionality.

    def savePolynomial(self, filename):
        """
        Saves the polynomial as an xml file.

        :param Poly self:
            An instance of the Poly class.
        :param string filename:
            Path location and filename.

        """
        root = tree.Element("polynomial")
        polydoc = tree.SubElement(root, "equadratures")
        tree.SubElement(polydoc, "version_number").text = str(VERSION_NUMBER)
        tree.SubElement(polydoc, "date").text = str(time.strftime("%d/%m/%Y"))

        basisinfo = 

        polydoc_basis = tree.SubElement(root, "multi-index")
        tree.SubElement(polydoc_basis, "basis_type").text = str(self.basis.basis_type)
        tree.SubElement(polydoc_basis, "maximum_orders").text = str(self.basis.orders)

        polydoc_coefficients = tree.SubElement(root, "coefficients")
        tree.SubElement(polydoc_coefficients, "coefficients").text = str(self.coefficients)
        tree.SubElement(polydoc_coefficients, "index_set").text = str(self.basis.elements)

        polynomialobject = tree.ElementTree(root)
        polynomialobject.write(filename, pretty_print=True)

Poly should facilitate rapid extrapolation. Rather than use

X2 = np.linspace(-1, 1., 100).reshape(100,1)
X = np.random.rand(20, 1)
y = X**2
p = Parameter(distribution='uniform', lower=-1., upper=1., order=3)
basis = Basis('univariate')
poly = Poly(p, basis, method='least-squares', sampling_args={'sample-points': X, 'sample-outputs': y})
poly.set_model()
f = poly.get_polyfit(X2)

It would simplfy workflows if method='least-squares' could be assumed by default, and the values in sampling_args could be provided as inputs permitting the following to work. Checks on the shape of X and y must ensure that the number of rows in X is equivalent to the number of rows in y and that y only has 1 column.

Also note that in the above the code should be OK with sampling_args={'sample_points': X, 'sample_outputs': y}

1. effectivequads.py
2. integration_utils & pseudospectral_utils.py
3. Comment qr.py
4. Check least squares with constraints implementation

Using PolyTree with the following parameters returns an error when logging=True:

order = 1
max_depth = 3

polytree = PolyTree(splitting_criterion='model_agnostic', max_depth=max_depth, min_samples_leaf=None, 
         k=0.5, order=order, basis='hyperbolic-basis', search='grid',samples=50,
         logging=True, poly_method='least-squares', poly_solver_args=None)
polytree.fit(X_train_eq,y_train.to_numpy())

The relevant error output is:

in _splitter(node)
    192                                                         # Update best parameters if loss is lower
    193                                                         if loss_split < loss_best:
--> 194                                                                 if self.logging: self.log.append({'event': 'best_split', 'data': {'j_feature':j_feature, 'threshold':threshold, 'loss': loss_split, 'poly_left': poly_left, 'poly_right': poly_right}})
    195                                                                 did_split = True
    196                                                                 loss_best = loss_split

UnboundLocalError: local variable 'poly_left' referenced before assignment

Can you double check the vegetation input parameter space ?

27 [ 1.44300000e+02, 2.35000000e-01, 6.00000000e-03, 6.00000000e-04],
28 [ 6.21152934e+01, 1.74581460e-01, 2.90161332e-03, 6.00000000e-04],
29 [ 2.26484707e+02, 2.95418540e-01, 6.00000000e-03, 2.90161332e-04],
30 [ 1.44300000e+02, 1.74581460e-01, 9.09838668e-03, 9.09838668e-04],
31 [ 1.44300000e+02, 2.95418540e-01, 2.90161332e-03, 9.09838668e-04],
32 [ 6.21152934e+01, 2.95418540e-01, 6.00000000e-03, 2.90161332e-04],
33 [ 2.26484707e+02, 1.74581460e-01, 6.00000000e-03, 2.90161332e-04],
34 [ 1.44300000e+02, 2.95418540e-01, 9.09838668e-03, 9.09838668e-04],
35 [ 1.44300000e+02, 2.35000000e-01, 9.09838668e-03, 2.90161332e-04],
36 [ 2.26484707e+02, 2.35000000e-01, 6.00000000e-03, 9.09838668e-04],
37 [ 1.44300000e+02, 2.35000000e-01, 2.90161332e-03, 2.90161332e-04],
38 [ 6.21152934e+01, 2.35000000e-01, 6.00000000e-03, 9.09838668e-04],
39 [ 6.21152934e+01, 1.74581460e-01, 6.00000000e-03, 2.90161332e-04],
40 [ 6.21152934e+01, 2.35000000e-01, 9.09838668e-03, 6.00000000e-04],
41 [ 1.44300000e+02, 1.74581460e-01, 2.90161332e-03, 9.09838668e-04]])

On newer versions of numpy, the attribute asscalar has been deprecated (in version 1.16) and removed as of version 1.23.

Older versions will trigger the following error:

AttributeError: module 'numpy' has no attribute 'asscalar'

Solution should be to switch to numpy.ndarray.item() instead.

Lines to fix:

Parameter should have a default uniform distribution and a default order of 1.

p = Parameter(lower=-1., upper=1.)
# TypeError: __init__() missing 2 required positional arguments: 'order' and 'distribution'

Parameter should take in a name in its constructor. If a name is not provided a default value of 'parameter' can be assigned.

horsepower = Parameter(name='horsepower', lower=180, upper=350)

Failing due to cvxpy/numpy depency mismatch. Full error below:

$ pip install cvxpy>=1.1
Could not find a version that satisfies the requirement numpy==1.19 (from versions: 1.9.0, 1.9.1, 1.9.2, 1.9.3, 1.10.0, 1.10.1, 1.10.2, 1.10.3, 1.10.4, 1.11.0, 1.11.1, 1.11.2, 1.11.3, 1.12.0, 1.12.1, 1.13.0rc1, 1.13.0rc2, 1.13.0, 1.13.1, 1.13.3, 1.14.0rc1, 1.14.0, 1.14.1, 1.14.2, 1.14.3, 1.14.4, 1.14.5, 1.14.6, 1.15.0rc1, 1.15.0rc2, 1.15.0, 1.15.1, 1.15.2, 1.15.3, 1.15.4, 1.16.0rc1, 1.16.0rc2, 1.16.0, 1.16.1, 1.16.2, 1.16.3, 1.16.4, 1.16.5, 1.16.6, 1.17.0rc1, 1.17.0rc2, 1.17.0, 1.17.1, 1.17.2, 1.17.3, 1.17.4, 1.17.5, 1.18.0rc1, 1.18.0, 1.18.1, 1.18.2, 1.18.3, 1.18.4, 1.18.5)
No matching distribution found for numpy==1.19
The command "pip install cvxpy>=1.1" failed and exited with 1 during .

This is a feature request for adding another class of distributions for equadratures.distributions.

Supporting Distributions like:

• Binomial Distribution
• Poisson Distribution
• Multinomial Distribution

Would work on this if approved.

This could be an issue throughout the repository. However, I have only tried to import parameter.py , hence encountered this issue here.

Python Version:

Encountered Error:

Expected Behaviour:

Reproduction steps:

In python console, type

from parameter import Parameter

The error above should be encountered,
Note that doing

from .parameter import Parameter

Should simply give

File "", line 1
import .parameter as Parameter
^
SyntaxError: invalid syntax

Double check and make sure "q" parameter can be given as an input.

Current pivoting technique is memory intensive. Need to investigate other techniques

def qrColumnPivoting_mgs(A):