Ask Your Question
1

derivative of MPolynomial_polydict

asked 2014-04-29 02:27:25 +0100

k1 gravatar image

updated 2015-01-14 16:21:58 +0100

FrédéricC gravatar image

In a for loop I generate a list of functions, then I take their derivatives with respect to different variables. For example, in the follwoing, Y[0,0] is a 'MPolynomial_polydict', and is equal to

x_0_1*x_0_1 + x_0_0*x_1_1 + x_0_2*x_0_2 + x_1_2*x_1_2 + x_0_0*x_2_2 + x_1_1*x_2_2 - 1.00000000000000

and list_of_variables[0] = x_0_1. When I take the derivative of the mentioned polynomial with respect to the mentioned variable I should get 2*x_0_1, but instead I get:

sage: derivative(Y[0,0],list_of_variables[0])
x_0_1

When I try this manually, it gives me the correct answer though:

sage: derivative(x_0_1*x_0_1 + x_0_0*x_1_1 + x_0_2*x_0_2 + x_1_2*x_1_2 + x_0_0*x_2_2 + x_1_1*x_2_2 - 1.00000000000000,x_0_1)
2*x_0_1

So, for some reason it considers the two x_0_1's in the first term of my polynomial as independent variables when I construct them automatically. Any ideas on why this problem arises, and how to fix this?

Edit:

This is how I build my vaiables:

names = [ [[] for i in range(n)] for j in range(n) ]

for i in range(n):
    for j in range(i+1):
        names[i][j] = (SR('x_' + str(j) + '_' + str(i)))
        names[j][i] = (SR('x_' + str(j) + '_' + str(i)))

R = PolynomialRing(CC,names[i][j] for i in range(n) for j in range(n)])
R.gens()
R.inject_variables()

More edit: Here is a the complete code:

#########################################################################
# Build variables, and the matrix corresponding to it
#########################################################################

def build_variables(n):
    names = [ [[] for i in range(n)] for j in range(n) ]

    for i in range(n):
        for j in range(i+1):
            names[i][j] = (SR('x_' + str(j) + '_' + str(i)))
            names[j][i] = (SR('x_' + str(j) + '_' + str(i)))

    return(names)


#########################################################################
# Define the function f that maps a matrix to the coefficients of its characteristic polynomial
#########################################################################
def CharPoly(Mat):
    X = matrix(Mat)
    n = X.ncols()
    C_X = X.characteristic_polynomial()
    Y = []
    for i in range(n):
        Y.append(C_X[i])
    return(Y)


############################################################################
# This solves that lambda problem
############################################################################
def lambda_siep(G,L,iter=100,epsilon = .1):
                        # G is any graph on n vertices
                        # L is the list of n desired distinct eigenvalues
                        # m is the number of itterations of the Newton's method
                        # epsilon: the off-diagonal entries will be equal to epsilon

    n = G.order()
    my_variables = build_variables(n)

    R = PolynomialRing(CC,[my_variables[i][j] for i in range(n) for j in range(n)])
    R.gens()
    R.inject_variables()


    X = [ [ R.gens()[n*i+j] for i in range(n) ] for j in range(n) ]

    Y = matrix(CharPoly(X)) - matrix(CharPoly(diagonal_matrix(L)))

    J = matrix(R,n)
    for i in range(n):
        for j in range(n):
            J[i,j] = derivative(Y[0][i],my_variables[j][j])

    B = diagonal_matrix(L) + epsilon * G.adjacency_matrix()

    count = 0
    while count < iter:

        T = [ B[i,j] for i in range(n) for j in range(n)]

        C = (J(T)).solve_right(Y(T).transpose())
        LC = list(C)

        B = B - diagonal_matrix([LC[i][0] for i in range(n)])

        count = count + 1

    return(B)
edit retag flag offensive close merge delete

Comments

I cannot reproduce your problem. Please give a complete working example. How much is `n`? Who's `Y`? There's a missing braket in your definition of `R`.

Luca gravatar imageLuca ( 2014-04-29 10:37:24 +0100 )edit

I included the complete code at the end of the post.

k1 gravatar imagek1 ( 2014-04-29 12:45:31 +0100 )edit

I'm sorry. This doesn't help either. If you want useful help, you must make an useful effort to isolate the potential bug. It is by putting less code, not more, that you will achieve this: I cannot guess what parameters to `lambda_siep` are going to trigger the bug.

Luca gravatar imageLuca ( 2014-04-30 12:12:04 +0100 )edit

1 Answer

Sort by » oldest newest most voted
1

answered 2014-07-25 10:46:11 +0100

FrédéricC gravatar image

There are several little problems in your code.

First you do not need to use SR in build_variables (avoid SR as much as possible when using polynomials). You can do

def build_variables(n):
    names = []

    for i in range(n):
        for j in range(n):
            names += ['x_' + str(i) + '_' + str(j)]

    return(names)

Then in the lambda_siep function, you should only use the result my_variables of build_variables once, inside the PolynomialRing constructor.

Then you should not use my_variables anymore, but instead your X in the line

J[i,j] = derivative(Y[0][i], X[j][j])

The two lines

R.gens()
R.inject_variables()

are useless and can be removed. In particular, inject_variables is only to be used in interactive mode.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

Stats

Asked: 2014-04-29 02:27:25 +0100

Seen: 489 times

Last updated: Jul 25 '14