testin 1, 2

pyssp/modeling.py

@@ -1,6 +1,5 @@
 """Chapter 4 modeling algorithm implementations."""
 
-from numpy.linalg import inv
 import numpy as np
 import scipy as sp
 
@@ -21,7 +20,7 @@ def pade(x, p, q):
     # Linear difference matrix spanning the number of zeros
     Xq = X[q + 1:q + p + 1, 1:p + 1].copy()
     print(Xq.shape)
-    a = linalg.solve(-Xq, X[q + 1: q + p + 1, 0])
+    a = np.linalg.solve(-Xq, X[q + 1: q + p + 1, 0])
     # a(0) normalized to 1
     a = np.concatenate((np.ones(1), a)).reshape(-1, 1)
     b = X[:q + 1, :p + 1] @ a
@@ -60,7 +59,7 @@ def prony(x, p, q):
     Xq1 = X[q + 1:N + p, 0].copy()
     Xq_H = Xq.conjugate().transpose()
     rx = Xq_H @ Xq1
-    Xinv = linalg.inv(Xq_H @ Xq)
+    Xinv = np.linalg.inv(Xq_H @ Xq)
     a = -Xinv @ rx
     print(a.shape)
     # a(0) normalized to 1
@@ -94,7 +93,7 @@ def shanks(x, p, q):
     gx = G0_H @ x0
     # the factorization does not guarantee nonsingularity!
     # resulting matrix is positive *semi*-definite
-    Ginv = linalg.inv(G0_H @ G0)
+    Ginv = np.linalg.inv(G0_H @ G0)
     print(f"{x.shape=}")
     print(f"{Ginv.shape=}")
     b = Ginv @ gx
@@ -118,7 +117,7 @@ def spike(g, n0, n):
     G_H = np.transpose(G.conjugate())
 
     print(f"{G_H.shape=}, {G.shape=}")
-    Ginv = linalg.inv(G_H @ G)
+    Ginv = np.linalg.inv(G_H @ G)
     h = Ginv @ G_H @ d
 
     return h
@@ -162,7 +161,7 @@ def covm(x, p):
     cx = X[p:N, 0].copy()
     Xq_H = Xq.copy().conjugate().transpose()
     print(f"{Xq=}")
-    Xinv = linalg.inv(Xq_H @ Xq)
+    Xinv = np.linalg.inv(Xq_H @ Xq)
     a1 = -Xinv @ Xq_H @ cx
     a = np.concatenate((np.ones(1), a1)).reshape(-1, 1)
     err = np.abs(cx.transpose() @ X[p:N,] @ a)