Software
Numerical range
Matlab / Octave
- An Effective Algorithm for Computing the Numerical Range by Carl C. Cowen and Elad Harel
- Field of values and numerical abscissa by Nick Trefethen, November
Mathematica
Python / Numpy
import numpy as np
import matplotlib.mlab as mlab
def numerical_range(A,resolution=0.01):
"""
Function implements algorithm for calculation of numerical range
http://www.math.iupui.edu/~ccowen/Downloads/33NumRange.html
"""
A=np.asmatrix(A)
th=np.arange(0,2*np.pi+resolution,resolution)
k=0
w=[]
for j in th:
Ath=np.exp(1j*-j)*A
Hth=(Ath+Ath.H)/2
e,r=np.linalg.eigh(Hth)
r=np.matrix(r)
e=np.real(e)
m=e.max()
s=mlab.find(e==m)
if np.size(s)==1:
w.append(np.matrix.item(r[:,s].H*A*r[:,s]))
else:
Kth=1j*(Hth-Ath)
pKp=r[:,s].H*Kth*r[:,s]
ee,rr=np.linalg.eigh(pKp)
rr=np.matrix(rr)
ee=np.real(ee)
mm=ee.min()
sm=mlab.find(ee==mm)
temp=rr[:,sm[0]].H*r[:,s].H*A*r[:,s]*rr[:,sm[0]]
w.append(temp[0,0])
k+=1
mM=ee.max()
sM=mlab.find(ee==mM)
temp=rr[:,sM[0]].H*r[:,s].H*A*r[:,s]*rr[:,sM[0]]
w.append(temp[0,0])
k+=1
return w
Julia
using LinearAlgebra
##
function numericalrange(A,resolution=0.01)
th = (0:resolution:2π)
k = 0
w = ComplexF64[]
for i in th
ath = exp(-i*1im)*A
hth = (ath + ath')/2
f = eigen(Hermitian(hth))
r = f.values
e = f.vectors
m = maximum(r)
s = findall(x ->x == m,r)
if length(s)==1
push!(w,(e[:,s]'*A*e[:,s])[1,1])
else
kth = 1im*(hth-ath)
pkp = e[:,s]'*kth*e[:,s]
ff = eigen(Hermitian(pkp))
rr = ff.values
ee = ff.vectors
mm = minimum(rr)
ss = findall(x ->x == mm,rr)
temp = ee[:,ss[1]]'*e[:,s]'*A*e[:,s]*ee[:,ss[1]]
push!(w,temp[1,1])
k+=1
mM = maximum(rr)
sM = findall(x ->x == mM,rr)
temp = ee[:,sM[1]]'*e[:,s]*A*e[:,s]*ee[:,sM[1]]
push!(w,temp[1,1])
k+=1
end
end
return w
end