Is there a way in cvxpy to have a conditional constraint, I am looking at a simple convex portfolio optimization problem like this one.
from cvxpy import *
import numpy as np
np.random.seed(1)
n = 10
Sigma = np.random.randn(n, n)
Sigma = Sigma.T.dot(Sigma)
w = Variable(n)
mu = np.abs(np.random.randn(n, 1))
ret = mu.T*w
risk = quad_form(w, Sigma)
orig_w = [0.15,0.2,0.2,0.2,0.2,0.05,0.0,0.0,0.0,0.0]
lambda_ret = Parameter(sign='positive')
lambda_ret = 5
lambda_risk = Parameter(sign='positive')
lambda_risk = 1
constraints = [sum_entries(w) == 1, w >= 0]
prob = Problem(Maximize(lambda_ret * ret - lambda_risk * risk ),constraints)
prob.solve()
and I am trying to introduce a constraint that will only apply to certain scenarios
sum_entries([ w[i]-orig_w[i] if w[i]-orig_w[i] >= 0 else 0 for i in range(n)]) >= some threshold
In this python pseudo code I would like to only control positive weight changes.
I looked through the cvxpy functions but nothing seemed to be able to do that.
In general, often manual formulation and integer-programming approaches are needed. There is no support for if-else descriptions in any modelling-tool i know of.
Your constraint:
cvx.sum_entries([ w[i]-orig_w[i] if w[i]-orig_w[i] >= 0 else 0 for i in range(n)]) >= some threshold
is equivalent to:
cvx.sum_entries([ cvx.pos(w[i]-orig_w[i]) for i in range(n)]) >= some threshold
This uses cvxpy's function pos:
pos(x) <-> max{x,0} x inR
(I did not check DCP-compatibility nor syntax; take this as general hint)
Remark:
cvx.sum_entries([ cvx.pos(w[i]-orig_w[i]) for i in range(n)]) >= some threshold
looks pretty unnatural to what you would usually do in cvxpy:
cvx.sum_entries(cvx.pos(w-orig_w)) >= some threshold
(vectorized form: much more performant and usually more clear)