When do I need to use: from future import division ?

Hi,

I'm trying to run a simple proof-of-concept entropy optimization in cvxpy:

import cvxpy as cp
import cvxopt
import numpy as np

N = 50
p = cp.Variable(N)
objective = cp.Maximize( sum(cp.entr(p)) )
cvx_constraints = [ np.ones( (1, N) ) * p == 1 ]
problem = cp.Problem(objective, cvx_constraints)
result = problem.solve()
print p.value

If N = 20, it finishes in < 1 second. If N = 30, it takes 23 seconds. If N = 50, it fails for some reason after 30 seconds (with no error) and prints None for p.value.

By contrast, I can run the identical MATLAB program with N = 1000 and it finishes successfully in < 2 minutes:

tic
N = 1000;
cvx_begin
    variable p(N)
    maximize(sum(entr(p)))
    subject to
        ones(1, N) * p == 1
cvx_end
toc

Is there anything to be done about this? Thanks!

The docs says "To check whether an Expression object follows the DCP rules, use the method expr.is_dcp(). Constraints, Objectives, and Problems also have an is_dcp method". However:

kbriggs:~/python> python
Python 2.7.5+ (default, Sep 19 2013, 13:48:49) [GCC 4.8.1] on linux2

import cvxpy as cp
x = cp.Variable(5)
x.is_dcp()
Traceback (most recent call last):
File "", line 1, in
AttributeError: 'Variable' object has no attribute 'is_dcp'
expr = 2*x # expr is an Expression object after each assignment.
expr.is_dcp()
Traceback (most recent call last):
File "", line 1, in
AttributeError: 'MulExpression' object has no attribute 'is_dcp'
type(expr)
<class 'cvxpy.atoms.affine.binary_operators.MulExpression'>

Hi,
I was trying to solve a mean-variance optimization problem with constraints on the weights to in [0,1]. The solver returns negative values for some of the weights. Here is the code:

import cvxpy as cp
import numpy as np
import cvxopt

n=3
muVector=np.matrix([7.2,9.8,2.5])
sigma = np.diag([100,50,10])

x=cp.Variable(n)

expectedReturn = muVector_x
risk = cp.quad_form(x,sigma)
riskaversion=(.5)__-20
objective=cp.Maximize(expectedReturn - riskaversion_risk)
constraints = [0 <= x, x <= 1,sum(x)==1]

p=cp.Problem(objective,constraints)
p.solve()
for w in x.value:
print w

-0.196307484923
-0.257650575912
1.45395806083

The code below (feel free to use it as a test case or example problem) calculates the maximum expected log return for investment in 10 different assets with 10 different outcome scenarios. The objective is similar to that of acent.py, but it is a weighted sum of logs. The solvers do not converge to the solution within maxiter, though it is a rather straightforward problem. Any ideas?

import cvxpy as cvx
import cvxopt

scenario_probabilities = cvx.numpy.array((0.3,0.15,0.1,0.1,0.06,0.06,0.06,0.06,0.06,0.05)).reshape(1,-1)
joint_returns = cvx.numpy.array( \
    [[ 1.09,  0.96,  0.95,  1.06,  1.09,  1.54,  1.16,  1.01,  0.82,  1.04], \
    [ 1.14,  1.02,  1.04,  1.07,  1.02,  0.93,  1.09,  0.85,  1.47,  1.21], \
    [ 1.01,  1.35,  0.95,  0.81,  1.05,  0.85,  1.02,  0.88,  1.  ,  1.05], \
    [ 1.1 ,  0.96,  1.13,  1.95,  1.06,  0.81,  1.02,  1.12,  0.95,  0.97], \
    [ 1.12,  0.9 ,  1.  ,  1.08,  1.1 ,  1.07,  1.7 ,  0.94,  0.97,  0.93], \
    [ 1.64,  1.01,  1.25,  0.97,  0.93,  1.03,  0.81,  1.13,  1.02,  0.91], \
    [ 0.85,  0.95,  0.98,  0.99,  1.1 ,  1.02,  1.04,  1.67,  0.96,  0.96], \
    [ 1.15,  0.97,  1.54,  0.95,  1.09,  1.14,  1.08,  0.9 ,  0.91,  0.82], \
    [ 1.13,  0.95,  1.1 ,  1.03,  1.89,  1.2 ,  0.97,  1.04,  1.04,  1.  ], \
    [ 0.93,  1.08,  0.84,  0.97,  0.95,  0.8 ,  0.95,  0.98,  1.29,  1.46]])

portfolio = cvx.Variable(10)
riskfree = cvx.Variable()
scenario_returns = riskfree + joint_returns.T * portfolio
objective = scenario_probabilities * cvx.log(scenario_returns) 
p = cvx.Problem(cvx.Maximize(objective),[scenario_returns >= 1e-5, riskfree + sum(portfolio) <= 1])

p.solve()

What is the plan for documentation?

Could I start making sphinx documentation?

Hey guys I don't know whether I did something wrong but when I installed cvxpy (already had cvxopt,numpy and scipy) it didn't let me run the first statement
"from cvxpy import * " because I didn't have the "ecos" module installed. So I would suggest that you either list that as one of the dependencies in the documentation or install it implicitly during cvxpy install. I see a .sh file that does install this but I don't think that is ever called, not a big deal just a suggestion. And it works fine after I installed "ecos" so thanks for your hard work!

How feasible would it be to add support for python 3? A lot of the solver interfaces and other packages these days are compatible, and quite a few people (including me) like to work in python 3 if possible.

Installation runs smoothly but there are some import issues when trying to load the cvxpy package. Seems to have something to do with namespaces?

When using numpy instead of cvxopt for A,

objective1 = Minimize( norm(A*x1-b, 1) )
File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/cvxpy/atoms/norm.py", line 29, in norm
x = Expression.cast_to_const(x)
File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/cvxpy/expressions/expression.py", line 114, in cast_to_const
return expr if isinstance(expr, Expression) else types.constant()(expr)
File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/cvxpy/expressions/constants/constant.py", line 33, in init
self.set_context()
File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/cvxpy/expressions/constants/constant.py", line 50, in set_context
sign = intf.sign(self.value)
File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/cvxpy/interface/matrix_utilities.py", line 98, in sign
mat = INTERFACES[np.ndarray].const_to_matrix(constant)
File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/cvxpy/interface/numpy_interface/ndarray_interface.py", line 32, in const_to_matrix
mat = numpy.array(value, dtype='float64')
ValueError: setting an array element with a sequence.

There's a comment in setup.py "this doesn't appear to do anything...".

Perhaps this page is helpful:

http://pythonhosted.org/setuptools/setuptools.html#declaring-dependencies

Maybe the 'install_requires' keyword should be used instead.

kbriggs:~/Downloads/cvxpy-master/examples> python qcqp.py
[ 4.69e+00 3.73e+00 3.66e+00 -4.17e+00 -3.56e+00 -2.45e-01]
[ 3.73e+00 8.01e+00 3.37e+00 -6.34e-01 -3.51e-01 -3.17e+00]
[ 3.66e+00 3.37e+00 6.14e+00 -5.39e+00 -4.22e+00 -7.17e-01]
[-4.17e+00 -6.34e-01 -5.39e+00 7.85e+00 5.11e+00 -6.85e-01]
[-3.56e+00 -3.51e-01 -4.22e+00 5.11e+00 7.31e+00 -1.02e+00]
[-2.45e-01 -3.17e+00 -7.17e-01 -6.85e-01 -1.02e+00 2.19e+00]

Traceback (most recent call last):
File "qcqp.py", line 65, in
if get_status(primal_result) is SOLVED:
NameError: name 'get_status' is not defined

Hi, I was wondering if cvxpy can handle problem with complex numbers.
I want to solve a optimazation problem like
minimize norm2(x)
subject to: Ax = b

in which A, b contains complex elements.
How should I program by CVXPY/
Thank you for helps.

I'm 90% sure the error is happening because I forgot to commit a file in my patch yesterday. I'll fix it as soon as I have computer access today.

This issue relates to question 4.1 of the EE364A additional exercises.

I tried this both on Matlab and Python and obtained different results for the dual variable, for the same problem and data. The primal optimal variable value is the same. The difference in dual variables between CVX for Matlab and CVXPY is on the order of 1e-1.

The following code snippets are the minimum necessary to see the differing dual variable values (lambda). I can provide the full scripts if that would be helpful.

A = [1 -0.5; -0.5 2];
C = [1 2; 1 -4];
E = [5 76];
d = [-2; -3];
f = 1;

cvx_begin quiet
    variable x(2)
    dual variable lambda1
    dual variable lambda2
    minimize quad_form(x, A) - x(1)
    subject to
        lambda1 : C*x <= d
        lambda2 : E*x <= f
cvx_end

optimal = cvx_optval
x
lambda = [lambda1; lambda2]

Matlab output:

>> add4q1_test
optimal =
    8.2222
x =
   -2.3333
    0.1667
lambda =
    1.8994
    3.4684
    0.0931

Python:

import cvxpy as cp
import numpy.matlib as ml
import itertools
mat = ml.asmatrix
concat = ml.concatenate

x = cp.Variable(2)
u = cp.Parameter(2)

A = mat('1 -0.5; -0.5 2')
C = mat('1 2; 1 -4')
E = mat('5 76')
f = 1
objective = cp.Minimize(cp.quad_form(x, A) - x[0])
constraints = [
    C*x <= u,
    E*x <= f
]
problem = cp.Problem(objective, constraints)

u_original = mat('-2; -3')
u.value = u_original

result = problem.solve()
status = cp.get_status(result)

x_val = mat(x.value)
lambda_val = concat((mat(constraints[0].dual_value), mat(constraints[1].dual_value)))

print "Status:", status
print "Optimal value p*:", result
print "Optimal primal value x*:", x_val.T
print "Optimal dual value lambda*:", lambda_val.T

Python output:

$ python add4q1_test.py
Status: solved
Optimal value p*: 8.2222216673
Optimal primal value x*: [[-2.33333352  0.16666658]]
Optimal dual value lambda*: [[ 1.18646785  3.95896591  0.13766984]]

This occurs in the current image_processing example.

If g is a generator,

vstack(*g)

gives the AttributeError.

Hey,

I'm just starting to work with this library, and I've gotten a few problems working, which is great. However, I'm currently trying to find the minimum volume ellipsoid containing a finite set of points, and I'm having a hard time understanding how to express that objective using this library. I haven't found either the logdet(A) or the det(A)^(1/n) functions available as concave objectives for A = Variable(n,n). Is this available and I have just missed it, or is this objective planned to be implemented?

Is there a way to silence the output to the console? The old syntax from 0.0.1 is not working any more. Thanks!

When creating a semidefinite cone, a MemoryError exception is thrown. It appears to be caused by using line 55 in semidefinite_cone.py. When m is assigned el.shape[0] this has a type of numpy.int32. Using this type to index G in line 65 causes the memory error. Changing line 55 to:

m = int(el.shape[0])

eliminated the exception.

Any chance there will be an elementwise entropy function coming to cvxpy? The MATLAB version of cvx has a function called entr, where entr(x) = x .* log(x). See http://cvxr.com/cvx/doc/funcref.html#built-in-functions.

I know I can construct this using the kl_div atom, but that takes scalar arguments so I have to add all the scalars in my (long) vector variable, which seems like it might be inefficient (?)

Hi,
This happened after the very last update. Now when I call prob.solve(), it prints a lot of stuff to the console. I am guessing it is due to something that was turned on for debugging. Anyway, just wanted to give you a quick feedback on that.

The code:

square(xi - yi for xi, yi in zip(expression_list, constant_list))

yields the error

Exception: <type 'generator'> is not a valid type for a Constant value.

Is there a way to model the log function? I believe it exists in CVXOPT, but I couldn't find it in CVXPY.

Hi,
In the old cvxpy, there is kl_div function. I couldn't find it in the new version. Is there something similar to kl_div ?
Thanks

working with sparse matrices results in an error, as evidenced in the 'stock_tradeoff.py' example script:

python examples/stock_tradeoff.py
Traceback (most recent call last):
File "stock_tradeoff.py", line 19, in
variance = square(norm2(F.T_x)) + square(norm2(D_x))
File "/python2.7/site-packages/cvxpy/expressions/expression.py", line 144, in rmul
return Expression.cast_to_const(other) * self
File "/python2.7/site-packages/cvxpy/expressions/expression.py", line 75, in cast_to_const
return expr if isinstance(expr, Expression) else types.constant()(expr)
File "/python2.7/site-packages/cvxpy/expressions/constant.py", line 34, in init
self.set_sign_curv()
File "/python2.7/site-packages/cvxpy/expressions/constant.py", line 54, in set_sign_curv
sign = intf.sign(self.value)
File "/python2.7/site-packages/cvxpy/interface/matrix_utilities.py", line 90, in sign
mat = NDARRAY_INTERFACE.const_to_matrix(constant)
File "/python2.7/site-packages/cvxpy/interface/numpy_interface/ndarray_interface.py", line 34, in const_to_matrix
return numpy.array(value, dtype='float64')
TypeError: float() argument must be a string or a number

If the PyPi release is updated, then people should be able to simply easy_install cvxpy or pip install cvxpy and get the dependencies automatically.

But apparently, easy_install cvxopt doesn't successfully install cvxopt on Windows, so I suppose automatic installation of dependencies for cvxpy isn't quite as useful.

Hi, I am writng a code for example 1 in
http://see.stanford.edu/materials/lsocoee364a/hw6sol.pdf
The probelm require k = 201, and I encountered solver_error during optimzation.
I tried k =40, then the result is perfected fitted. Is this probelm related to the scalibility of cvxpy? (The code is pasted here)

import numpy as np
import cvxpy as cp
k = 80
t = [-3.0+6.0*(i)/(k-1) for i in range(k) ]
y = np.exp(t)
T_powers = np.matrix(np.hstack((np.ones((k,1)),np.matrix(t).T,np.power(np.matrix(t).T,2))))
u = np.exp(3)
l = 0
bisection_tol = 1e-3
gamma1 = cp.Parameter(sign='positive')
a = cp.Variable(3)
b = cp.Variable(2)
objective = 0

constraints = [cp.abs(T_powers[i]a-y[i](T_powers[i]cp.vstack(1,b)))<=gamma1(T_powers[i]*cp.vstack(1,b)) for i in range(100)]

constraints = [T_powers_a-np.diag(y)(T_powers_cp.vstack(1,b))<=gamma1(T_powers*cp.vstack(1,b)),

T_powers_a-np.diag(y)(T_powers_cp.vstack(1,b))>=-gamma1(T_powers_cp.vstack(1,b))]

constraints = [cp.abs(T_powers_a-np.diag(y)(T_powers_cp.vstack(1,b)))<=gamma1(T_powers_cp.vstack(1,b))]
objective = cp.Minimize(0)
p = cp.Problem(objective,constraints)
gamma1.value = (l+u)/2.0
a_opt = 0
b_opt = 0
gamma1.value = (u+l)/2
while (u-l)>=bisection_tol:
print p.is_dcp()
p.solve()
if p.status is 'optimal':
u = gamma1.value
a_opt = a.value
b_opt = b.value
#print a_opt
#print b_opt
objval_opt = gamma1.value
#print 'here'

else:
    l = gamma1.value
gamma1.value = (l+u)/2
#print gamma1.value
print p.status

y_fit = T_powers*a_opt

print a_opt
print b_opt
print gamma1.value
plot(t,y,'b')
f_fit = np.divide(T_powers_a_opt,T_powers_np.vstack((1,b_opt)))
plot(t,f_fit,'r+')

I am looking for a GP example.

running "nosetests" in the base directory yields an error:

ERROR: Failure: ImportError (No module named containers)
Traceback (most recent call last):
File "/Library/Python/2.7/site-packages/nose-1.2.1-py2.7.egg/nose/loader.py", line 390, in loadTestsFromName
addr.filename, addr.module)
File "/Library/Python/2.7/site-packages/nose-1.2.1-py2.7.egg/nose/importer.py", line 39, in importFromPath
return self.importFromDir(dir_path, fqname)
File "/Library/Python/2.7/site-packages/nose-1.2.1-py2.7.egg/nose/importer.py", line 86, in importFromDir
mod = load_module(part_fqname, fh, filename, desc)
File "/Users/echu/src/cvxpy/tests/test_expressions.py", line 5, in
from cvxpy.expressions.containers import Variables
ImportError: No module named containers

as much as "from cvxpy import *" is convenient, i think we should either

(a) discourage this or
(b) ensure that our atoms escape to builtins when appropriate

i was using some code that involved "min" somewhere and it just did not work as expected; the error messages are very cryptic and the result is actually quite surprising.

changing import to 'cvxpy.atoms.vstack' is most obvious fix, but that creates a circular dependency in the Atom class.

Traceback (most recent call last):
File "app.py", line 4, in
from cvxpy import *
File "/python2.7/site-packages/cvxpy/init.py", line 20, in
from atoms import *
File "/python2.7/site-packages/cvxpy/atoms/init.py", line 20, in
from abs import abs
File "/python2.7/site-packages/cvxpy/atoms/abs.py", line 20, in
from atom import Atom
File "/python2.7/site-packages/cvxpy/atoms/atom.py", line 21, in
from cvxpy.expressions.variable import Variable
File "/python2.7/site-packages/cvxpy/expressions/variable.py", line 21, in
import cvxpy.interface.matrix_utilities as intf
File "/python2.7/site-packages/cvxpy/interface/matrix_utilities.py", line 23, in
import cvxpy.utilities as u
File "/python2.7/site-packages/cvxpy/utilities/init.py", line 22, in
from curvature import Curvature
File "/python2.7/site-packages/cvxpy/utilities/curvature.py", line 21, in
from vstack import vstack
ImportError: No module named vstack

After installing cvxpy and running tests I got about 60 errors like
dims['l'] ought to be a nonnegative integer

For example this code works fine
import cvxopt
import cvxpy as cp
c = cvxopt.matrix([1, 1])
x = cp.Variable(2)
objective = cp.Minimize(c.T_x)
a1 = cvxopt.matrix([2, 1], (1, 2))
a2 = cvxopt.matrix([1, 3], (1, 2))
b1 = 1
b2 = 1
constraints = [0 <= x, a1_x >= b1, a2*x >= b2]
p = cp.Problem(objective, constraints)
result = p.solve()

This code:
import cvxopt
import cvxpy as cp
c = cvxopt.matrix([1, 1])
x = cp.Variable(2)
objective = cp.Minimize(c.T_x)
A = cvxopt.matrix([2, 1, 1, 3], (2, 2))
b = cvxopt.matrix([1, 1])
constraints = [0 <= x, A_x >= b]
p = cp.Problem(objective, constraints)
result = p.solve()
raise error
dims['l'] ought to be a nonnegative integer

QP problem from examples like this
n = 3
P = cvxopt.matrix([13, 12, -2, 12, 17, 6, -2, 6, 12], (n,n))
q = cvxopt.matrix([-22, -14.5, 13], (n,1))
r = 1
x = Variable(n)
objective = Minimize( 0.5 * quad_form(x, P) + q.T * x + r )
constraints = [ x >= -1, x <= 1]
p = Problem(objective, constraints)
result = p.solve()
raise error
dims['q'] ought to be a list of positive integers

Hi,

Can you tell me if Variable can be sliced and how. I try the following code
x= Variable(3,30)
x[:,1]
it gave Exception: Invalid indices 1,slice(2, 2, None) for 'var30194'.

Thank you for any help!

Solving the following SDP problem, gives me a matrix that is not positive semi-definite. I was wondering what the problem is.

import numpy as np
from pylab import *
import cvxopt as co
import cvxpy as cp

x1 = cp.Variable()
x2 = cp.Variable()
x3 = cp.Variable()
t = cp.Variable()

A0 = co.matrix([[0,0,0,-2],[0,0,0,-3],[0,0,0,-4],[-2,-3,-4,0]])
A1 = co.matrix([[0,0,0,1],[0,0,0,2],[0,0,0,3],[1,2,3,0]])
A2 = co.matrix([[0,0,0,1],[0,0,0,1],[0,0,0,2],[1,1,2,0]])
A3 = co.matrix([[1,0,0,0],[0,1,0,0],[0,0,1,0],[0,0,0,1]])
I = np.identity(4)

objective = cp.Minimize(x3)

constraints = [ A0+A1_x1+A2_x2+A3*x3 >= 0 ]

p = cp.Problem(objective, constraints)

result=p.solve()
print(result)

print(x1.value)

print(x2.value)

print(x3.value)

print(np.linalg.eig(A0+A1_x1.value+A2_x2.value+A3*x3.value)[0])

I tried to use CVXPY with the CVXOPT solver, and ended up with the following exception:

ValueError: Rank(A) < p or Rank([G; A]) < n

It'd be nice if CVXPY could catch this exception and emit a more information error message about my constraints. I think a user shouldn't need to know what CVXOPT calls G.

When I run the following code in Sage, the problem is unbounded. I was wondering what the problem is.

from future import division
RealNumber=float
Integer=int
from cvxopt.base import matrix as m
import cvxopt
import numpy
from pylab import *
import math
from cvxpy import *

x = Variable(2)

constraints

a1 = m([1,1])
a2 = m([2,1])
constraints = [ a1.T_x >= -1, a2.T_x <= 10]

objective

objective = Maximize(sum(x))

p = Problem(objective, constraints)
result = p.solve()
print result

The optimal value

print x.value

The example at the head of README.md and the examples in the examples directory do not run. Inside Problem.solve(), SciPy throws a "ValueError: invalid shape" exception. The example notebook pa3p8 has the same problem, but pa13p3 runs correctly.

I am running Ubuntu Saucy (13.10), Python 2.7.5, IPython 1.2.0, CVXOPT 1.1.6, and the latest cvxpy from git.

I can't get figure out how to use this. Maybe a test or example could be included for the KL divergence of a variable distribution from a constant distribution (or constant from variable, or variable from variable)?

First of all, thanks for cvxpy. It seems to me like a promising and already quite useful piece of software.

I have noted that comparisons with numpy arrays of cvxpy Variable objects do not function. Minimal illustration follows (Python 2.7):

import numpy as np
import cvxpy as cp

a = np.array([cp.Variable(), cp.Variable()], dtype=object)
b = np.array([cp.Variable(), cp.Variable()], dtype=object)
c = 3

print a <= c
print a == c
print a > c
print a == b
print a >= b

The result of each comparison is [True, True], but I had hoped it would produce an object array with comparisons like array([var1 <= 3, var2 <=3]) etc.

I experience conflicts with pylab --inline in Ipython notebooks. In a nutshell:

%pylab inline
import cvxpy as cp

x = cp.Variable(5)
objective = cp.Minimize(sum(cp.square(x)))

throws the exception "type 'str' is not a valid type for a Constant value.". No problem without pylab inline. My installation is latest cvxpy from github repo, Python 2.7.3, Matplotlib 1.3.1, Ipython 1.0.0. and Windows 8 64bit. Thanks!

In examples/geometry/separating_hyperplanes.py, there is:

import convex_sets as cs

but it generates:
File "/../cvxpy-master/examples/geometry/convex_sets.py", line 4, in
from cvxpy.expressions.affine import AffObjective
ImportError: No module named affine

I was wondering what the problem is.

Hi,
I would like to know if there is any way to create objective function with log_normcdf like cvx?

Hi, Thank you for your excellent project. I am new to python and now trying to run an example from the text book of convex optimazation 4.17 (addition excercise from online course).
I would like to find minimum values for each row in a matrix containing expressions and then sum those min values (this will be served as objective). However, I recieved a buffer format not supported error. I guess min cannot take matrix containing expressions as arguments, then what should I do? Any helps is highly appreciated.

I wanted to make the schur complement of a positive semidefinite matrix in a way that cvxpy understands the schur complement properties.

Hi, I wonder if cvxpy can be used for modeling quadratic programing and how to fomulate it . For instance, is there something like quad_form in cvx? or Can I express qp into a equavelent form? (in matrix form)

Thank you very much

I generated a random 10 x 20 matrix Q and took P = Q.T * Q, so this is a positive semidefinite matrix. I then defined a variable x. Calling cvxpy.quad_form(x, P) fails with the exception "P has both positive and negative eigenvalues". Indeed, some of the zero eigenvalues are stored as -1e-14 due to numerical precision. However, for some reason, quad_form in the Matlab version of CVX handles such a P with no problem (even though it has the same numerical issues with negative eigenvalues).

kbriggs:~/Downloads/cvxpy-master/examples/geometry> python separating_polyhedra.py
Traceback (most recent call last):
File "separating_polyhedra.py", line 4, in
import convex_sets as cs
File "/home/kbriggs/Downloads/cvxpy-master/examples/geometry/convex_sets.py", line 4, in
from cvxpy.expressions.affine import AffObjective
ImportError: No module named affine

I am solving a linear program where objective is minimization is norm1. Clearly the objective is bounded. After adding a few constraints and solving it with cvxpy.ECOS, I get status as "unbounded". If I solve the same problem, without the constraints I get an answer. The data and math program are documented here:
http://nbviewer.ipython.org/gist/vijayvd/8319396

The latexed version of the math program can be seen if you download the ipython notebook. For some reason it is not rendering on the nbviewer site.

I have a log det sdp problem with linear inequality constraints in matrix variables. I tried to solve it using new cvxpy but it couldn't solve it. I got 'solver_error' in prog.status.

I tried to solve it using old cvxpy and at first, I got "Terminated (singular KKT matrix)" error message. Then I decreased the precision using
prog.options['abstol']=1e-02
prog.options['feastol']=1e-04
prog.options['reltol']=1e-04
and it worked fine.

How can we modify solver options using new cvxpy ? (I am using cvxopt's solver)

Hi all,
I tried installing cvxpy but when running setup.py I ran in the following error:

error: package directory 'cvxpy/atoms/nonlinear' does not exist

Indeed there is no nonlinear folder in the atoms directory, my solution was to comment out line 12 in setup.py in order to skip the package.

Currently, the "value" property of an IndexVariable accesses its parent's value.

It ought to check if the parent.value is None before attempting to index into it with self.key.

# The value at the index.
@property
def value(self):
    return self.parent.value[self.key]