Skip to content
/ ortho Public

A python package to generate orthogonal functions for regression

License

Notifications You must be signed in to change notification settings

ameli/ortho

Repository files navigation

ortho

A python package to generate a set of orthogonal functions.

Description

This package generates a set of orthonormal functions, called image01 , based on the set of non-orthonormal functions image02 defined by the inverse-monomials

image11

The orthonormalized functions image01 are the linear combination of the functions image02 by

image12

The functions image01 are orthonormal in the interval image03 with respect to the weight function image04 . That is,

image13

where image05 is the Kronecker delta function. The orthogonal functions are generated by Gram-Schmidt orthogonalization process. This script produces the symbolic functions using Sympy, a Python computer algebraic package. An application of these orthogonal functions can be found in [1].

Build and Test Status

codecov-devel

Platform Arch Python Version Continuous Integration
3.9 3.10 3.11 3.12
Linux X86-64 build-linux
AARCH-64
macOS X86-64 build-macos
ARM-64
Windows X86-64 build-windows
ARM-64

Install

Install using either of the following three methods.

1. Install from PyPi

pypi pyversions format

Install using the package available on PyPi by

pip install ortho

2. Install from Anaconda Cloud

conda-version conda-platform

Install using Anaconda cloud by

conda install -c s-ameli ortho

3. Install from Source Code

release licence implementation

Install directly from the source code by

git clone https://github.com/ameli/ortho.git
cd ortho
pip install .

Testing

To test the package, download the source code and use one of the following methods in the directory of the source code:

  • Method 1: test locally by:

    python setup.py test
    
  • Method 2: test in a virtual environment using tox:

    pip install tox
    tox
    

Usage

The package can be used in two ways:

1. Import as a Module

>>> from ortho import OrthogonalFunctions

>>> # Generate object of orthogonal functions
>>> OF = OrthogonalFunctions(
...        start_index=1,
...        num_func=9,
...        end_interval=1,
...        verbose=True)

>>> # Get numeric coefficients alpha[i] and a[i][j]
>>> alpha = OF.alpha
>>> a = OF.coeffs

>>> # Get symbolic coefficients alpha[i] and a[i][j]
>>> sym_alpha = OF.sym_alpha
>>> sym_a = OF.sym_coeffs

>>> # Get symbolic functions phi[i]
>>> sym_phi = OF.sym_phi

>>> # Print Functions
>>> OF.print()

>>> # Check mutual orthogonality of Functions
>>> status = OF.check(verbose=True)

>>> # Plot Functions
>>> OF.plot()

The parameters are:

  • start_index: the index of the starting function, image06 . Default is 1.
  • num_func: number of orthogonal functions to generate, image07 . Default is 9.
  • end_interval: the right interval of orthogonality, image08 . Default is 1.

2. Use As Standalone Application

The standalone application can be executed in the terminal in two ways:

  1. If you have installed the package, call ortho executable in terminal:

    ortho [options]
    

    The optional argument [options] will be explained in the next section. When the package ortho is installed, the executable ortho is located in the /bin directory of the python.

  2. Without installing the package, the main script of the package can be executed directly from the source code by

    # Download the package
    git clone https://github.com/ameli/ortho.git
    
    # Go to the package source directory
    cd ortho
    
    # Execute the main script of the package
    python -m ortho [options]
    

Optional arguments

When the standalone application (the second method in the above) is called, the executable accepts some optional arguments as follows.

Option Description
-h, --help Prints a help message.
-v, --version Prints version.
-l, --license Prints author info, citation and license.
-n, --num-func[=int] Number of orthogonal functions to generate. Positive integer. Default is 9.
-s, --start-func[=int] Starting function index. Non-negative integer. Default is 1.
-e, --end-interval[=float] End of the interval of functions domains. A real number greater than zero. Default is 1.
-c,--check Checks orthogonality of generated functions.
-p, --plot Plots generated functions, also saves the plot as pdf file in the current directory.

Parameters

The variables image06 , image07 , and image08 can be set in the script by the following arguments,

Variable Variable in script Option
image06 start_index -s, or --start-func
image07 num_func -n, or --num-func
image08 end_interval -e, or --end-interval

Examples

  1. Generate nine orthogonal functions from index 1 to 9 (defaults)

    ortho
    
  2. Generate eight orthogonal functions from index 1 to 8

    ortho -n 8
    
  3. Generate nine orthogonal functions from index 0 to 8

    ortho -s 0
    
  4. Generate nine orthogonal functions that are orthonormal in the interval [0,10]

    ortho -e 10
    
  5. Check orthogonality of each two functions, plot the orthonormal functions and save the plot to pdf

    ortho -c -p
    
  6. A complete example:

    ortho -n 9 -s 1 -e 1 -c -p
    

Output

  • Displays the orthogonal functions as computer algebraic symbolic functions. An example a set of generated functions is shown below.
phi_1(t) =  sqrt(x)
phi_2(t) =  sqrt(6)*(5*x**(1/3) - 6*sqrt(x))/3
phi_3(t) =  sqrt(2)*(21*x**(1/4) - 40*x**(1/3) + 20*sqrt(x))/2
phi_4(t) =  sqrt(10)*(84*x**(1/5) - 210*x**(1/4) + 175*x**(1/3) - 50*sqrt(x))/5
phi_5(t) =  sqrt(3)*(330*x**(1/6) - 1008*x**(1/5) + 1134*x**(1/4) - 560*x**(1/3) + 105*sqrt(x))/3
phi_6(t) =  sqrt(14)*(1287*x**(1/7) - 4620*x**(1/6) + 6468*x**(1/5) - 4410*x**(1/4) + 1470*x**(1/3) - 196*sqrt(x))/7
phi_7(t) =  5005*x**(1/8)/2 - 10296*x**(1/7) + 17160*x**(1/6) - 14784*x**(1/5) + 6930*x**(1/4) - 1680*x**(1/3) + 168*sqrt(x)
phi_8(t) =  sqrt(2)*(19448*x**(1/9) - 90090*x**(1/8) + 173745*x**(1/7) - 180180*x**(1/6) + 108108*x**(1/5) - 37422*x**(1/4) + 6930*x**(1/3) - 540*sqrt(x))/3
phi_9(t) =  sqrt(5)*(75582*x**(1/10) - 388960*x**(1/9) + 850850*x**(1/8) - 1029600*x**(1/7) + 750750*x**(1/6) - 336336*x**(1/5) + 90090*x**(1/4) - 13200*x**(1/3) + 825*sqrt(x))/5
  • Displays readable coefficients, image09 and image10 of the functions. For instance,
  i      alpha_i                                    a_[ij]
------  ----------   -----------------------------------------------------------------------
i = 1:  +sqrt(2/2)   [1                                                                    ]
i = 2:  -sqrt(2/3)   [6,   -5                                                              ]
i = 3:  +sqrt(2/4)   [20,  -40,    21                                                      ]
i = 4:  -sqrt(2/5)   [50,  -175,   210,   -84                                              ]
i = 5:  +sqrt(2/6)   [105, -560,   1134,  -1008,   330                                     ]
i = 6:  -sqrt(2/7)   [196, -1470,  4410,  -6468,   4620,   -1287                           ]
i = 7:  +sqrt(2/8)   [336, -3360,  13860, -29568,  34320,  -20592,   5005                  ]
i = 8:  -sqrt(2/9)   [540, -6930,  37422, -108108, 180180, -173745,  90090,  -19448        ]
i = 9:  +sqrt(2/10)  [825, -13200, 90090, -336336, 750750, -1029600, 850850, -388960, 75582]
  • Displays the matrix of the mutual inner product of functions to check orthogonality (using option -c). An example of the generated matrix of the mutual inner product of functions is shown below.
[[1 0 0 0 0 0 0 0 0]
 [0 1 0 0 0 0 0 0 0]
 [0 0 1 0 0 0 0 0 0]
 [0 0 0 1 0 0 0 0 0]
 [0 0 0 0 1 0 0 0 0]
 [0 0 0 0 0 1 0 0 0]
 [0 0 0 0 0 0 1 0 0]
 [0 0 0 0 0 0 0 1 0]
 [0 0 0 0 0 0 0 0 1]]
  • Plots the set of functions (using option -p) and saves the plot in the current directory. An example of a generated plot is shown below.

Citation

[1]Ameli, S., and Shadden. S. C. (2022). Interpolating Log-Determinant and Trace of the Powers of Matrix A + t B. Statistics and Computing 32, 108. paper-doi paper-arxiv
[2]Ameli, S. (2022). ameli/ortho: (v0.2.0). Zenodo. code-doi