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

doc: increased iterations in twist tutorial #187

Merged
merged 1 commit into from
Aug 23, 2024
Merged

doc: increased iterations in twist tutorial #187

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 7 additions & 7 deletions tutorials/twist.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,7 +59,7 @@ def callback(x, pf, pg, eps, cost):
L = np.abs((Aop.H * Aop).eigs(1)[0])
tau = 0.95 / L
eps = 5e-3
maxit = 100
maxit = 300

# IHT
l0 = pyproximal.proximal.L0Ball(3)
Expand Down Expand Up @@ -99,6 +99,7 @@ def callback(x, pf, pg, eps, cost):
callback=lambda x: callback(x, l2, l1, eps, costa))
nitera = len(costa)


# TWIST (Note that since the smallest eigenvalue is zero, we arbitrarily
# choose a small value for the solver to converge stably)
l1 = pyproximal.proximal.L1(sigma=eps)
Expand Down Expand Up @@ -135,11 +136,11 @@ def callback(x, pf, pg, eps, cost):
###############################################################################
# Finally, let's compare the converge behaviour of the different algorithms

fig, ax = plt.subplots(1, 1, figsize=(8, 3))
ax.semilogy(costi, 'r', lw=2, label=r'$x_{ISTA} (niter=%d)$' % niteri)
ax.semilogy(costf, 'g', lw=2, label=r'$x_{FISTA} (niter=%d)$' % niterf)
ax.semilogy(costa, 'm', lw=2, label=r'$x_{AA-ISTA} (niter=%d)$' % nitera)
ax.semilogy(costt, 'b', lw=2, label=r'$x_{TWIST} (niter=%d)$' % maxit)
fig, ax = plt.subplots(1, 1, figsize=(8, 4))
ax.loglog(costi, 'r', lw=2, label=r'$x_{ISTA} (niter=%d)$' % niteri)
ax.loglog(costf, 'g', lw=2, label=r'$x_{FISTA} (niter=%d)$' % niterf)
ax.loglog(costa, 'm', lw=2, label=r'$x_{AA-ISTA} (niter=%d)$' % nitera)
ax.loglog(costt, 'b', lw=2, label=r'$x_{TWIST} (niter=%d)$' % maxit)
ax.set_title('Cost function', size=15, fontweight='bold')
ax.set_xlabel('Iteration')
ax.legend()
Expand All @@ -154,4 +155,3 @@ def callback(x, pf, pg, eps, cost):
# When this information is not available, IHT may become suboptimal. In this case the
# FISTA or AA-ISTA solvers should always be preferred (over ISTA) and TwIST represents
# an alternative worth checking.

Loading