Class PLSR

source code

          object --+        
                   |        
 bases.BaseAlgorithm --+    
                       |    
 bases.ImplicitAlgorithm --+
                           |
              object --+   |
                       |   |
bases.IterativeAlgorithm --+
                           |
                          PLSR

A NIPALS implementation for PLS regresison.

Parameters
----------
max_iter : Non-negative integer. Maximum allowed number of iterations.
        Default is 200.

eps : Positive float. The tolerance used in the stopping criterion.

Examples
--------
>>> from parsimony.algorithms.nipals import PLSR
>>> import numpy as np
>>> np.random.seed(42)
>>>
>>> X = np.random.rand(10, 10)
>>> Y = np.random.rand(10, 5)
>>> w = np.random.rand(10, 1)
>>> c = np.random.rand(5, 1)
>>> plsr = PLSR()
>>> w, c = plsr.run([X, Y], [w, c])
>>> w
array([[ 0.34682103],
       [ 0.32576718],
       [ 0.28909788],
       [ 0.40036279],
       [ 0.32321038],
       [ 0.39060766],
       [ 0.22351433],
       [ 0.28643062],
       [ 0.29060872],
       [ 0.23712672]])
>>> c
array([[ 0.29443832],
       [ 0.35886751],
       [ 0.33847141],
       [ 0.23526002],
       [ 0.35910191]])
>>> C, S, W = np.linalg.svd(np.dot(Y.T, X))
>>> w_ = W[0, :].reshape(10, 1)
>>> w_ = -w_ if w_[0, 0] < 0.0 else w_
>>> w = -w if w[0, 0] < 0.0 else w
>>> round(np.linalg.norm(w - w_), 15)
1.52884e-10
>>> abs(np.dot(np.dot(X, w).T,
...            np.dot(Y, c / np.linalg.norm(c)))[0, 0] - S[0]) < 5e-15
True

A NIPALS implementation for PLS regresison.

Parameters
----------
XY : List of two numpy arrays. XY[0] is n-by-p and XY[1] is n-by-q. The
        independent and dependent variables.

wc : List of numpy array. The start vectors.

Returns
-------
w : Numpy array, p-by-1. The weight vector of X.

c : Numpy array, q-by-1. The weight vector of Y.