Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Solve quadprog #1896

Closed
wants to merge 14 commits into from
Closed

Solve quadprog #1896

Changes from 1 commit
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
5e97c16
compile but this always casts vars to doubles
rok-cesnovar Apr 16, 2020
6144ebd
remove the need for the second function
rok-cesnovar Apr 17, 2020
931fe66
solve_quadprog works for G is var
rok-cesnovar Apr 18, 2020
07a4932
more templating
rok-cesnovar Apr 18, 2020
4c3a855
any arg can be a parameter
rok-cesnovar Apr 18, 2020
04020e9
tests
rok-cesnovar Apr 18, 2020
c57e758
Merge remote-tracking branch 'origin/develop' into solve_quadprog
rok-cesnovar May 9, 2020
3a05585
added solve_quadprog_chol
rok-cesnovar May 9, 2020
ace2bdb
Merge remote-tracking branch 'origin/develop' into solve_quadprog
rok-cesnovar May 11, 2020
afe2c16
add Sean's optimizations
rok-cesnovar May 11, 2020
d6b6491
syntax fixes
rok-cesnovar May 11, 2020
f076923
fix compile issues
rok-cesnovar May 11, 2020
7b9f332
add Sean's corrections
rok-cesnovar May 18, 2020
78c3398
Smaller epsilon value
spinkney May 19, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Smaller epsilon value
  • Loading branch information
@spinkney
spinkney committed May 19, 2020
commit 78c33987dc5022337bb4e2dfeb44cb50d3189325
24 changes: 8 additions & 16 deletions stan/math/rev/fun/solve_quadprog.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -316,8 +316,9 @@ solve_quadprog(const Eigen::Matrix<T0, -1, -1> &G,
/* compute the inverse of the factorized matrix G^-1, this is the initial
* value for H */
//J = L^-T
Eigen::Matrix<T0, -1, -1> Li = (1e-9 < L.array().abs()).select(L, 0.0f);
auto J = stan::math::transpose(stan::math::mdivide_left_tri_low(Li));
Eigen::Matrix<T0, -1, -1> Li = (1e-14 < L.array().abs()).select(L, 0.0f);
auto J = stan::math::mdivide_left_tri_low(Li);
J.transposeInPlace();
auto c2 = J.trace();
#ifdef TRACE_SOLVER
print_matrix("J", J, n);
Expand Down Expand Up @@ -629,18 +630,8 @@ auto
*/

/* compute the trace of the original matrix G */
/* can remove this auto chol = L;
* can provide the trace already so don't need
* to compute G again
* auto G = tcrossprod(L);
* auto c1 = G.trace(); */

/* tr(AB^T) = sum( A hadamard_prod B)
* https://en.wikipedia.org/wiki/Trace_(linear_algebra)
*/
// auto c1 = elt_multiply(L, L).sum();
auto c1 = tcrossprod(L).trace();
/* initialize the matrix R */
/* initialize the matrix R */
d.setZero();

/* compute the trace of the original matrix G */
Expand All @@ -655,8 +646,9 @@ auto
/* **CHANGE HERE**
* J is only called once and trace is the same regardless of transpose
* also change chol to L */
Eigen::Matrix<T0, -1, -1> Li = (1e-9 < L.array().abs()).select(L, 0.0f);
auto J = stan::math::transpose(stan::math::mdivide_left_tri_low(Li));
Eigen::Matrix<T0, -1, -1> Li = (1e-14 < L.array().abs()).select(L, 0.0f);
auto J = stan::math::mdivide_left_tri_low(Li);
J.transposeInPlace();
auto c2 = J.trace();
// std::cout << c2 << "\n";
#ifdef TRACE_SOLVER
Expand Down Expand Up @@ -936,4 +928,4 @@ auto
}
}
}
#endif
#endif