Skip to content

Latest commit

 

History

History
57 lines (43 loc) · 1.02 KB

README.md

File metadata and controls

57 lines (43 loc) · 1.02 KB
Raw

NNQP

Efficient implementation of nonnegative quadratic programming solver in numba The solver implements a coordinate descent algorithm to find a solution to the problem

1/2 X^T Q X + f' X

subject to

X>=0

Authors: Andrea Giovannucci and Cengiz Pehlevan.

Installation

Requirements

  • numpy
  • scipy
  • numba (Anaconda distribution suggested)
  • cvxpy (only if you want to compare with another solver)

Example

import numpy as  np
import nnqp
#%% EXAMPLE WITH RANDOM VARIABLES
p = 10
Q1 = np.random.randn(p,p)
Q = Q1.dot(Q1.T)
f = np.random.randn(p)

#%% solve with NNQP
x = nnqp.nnqp(Q,f)
print x

#%% run only if you want to compare with another solver (requires cvxpy)
import cvxpy as cvx
cx = cvx.Variable(p) 
objective=cvx.Minimize(0.5*cvx.quad_form(cx, Q) + f.T*cx ) 
constraints= [cx >= 0]
prob = cvx.Problem(objective, constraints)
result = prob.solve(solver='SCS')
print(cx.value.T)

#%% compute the difference between outputs
print(np.linalg.norm(x-cx.value.T))/p