diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
index 5008449be4eb1c53ded405dc3a0c4558d79a9894..3a789f2add5cab2eadf8387a18a6e79ad57d825c 100644
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -119,3 +119,8 @@ run_curve_fitting:
artifacts:
paths:
- '*.png'
+
+run_visual_mgvi:
+ stage: demo_runs
+ script:
+ - python3 demos/mgvi_visualized.py
diff --git a/demos/mgvi_visualized.py b/demos/mgvi_visualized.py
new file mode 100644
index 0000000000000000000000000000000000000000..d8c06463bb5dd4aebf621fae9092afaec301c876
--- /dev/null
+++ b/demos/mgvi_visualized.py
@@ -0,0 +1,67 @@
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+# Copyright(C) 2013-2020 Max-Planck-Society
+# Authors: Reimar Leike, Philipp Arras
+#
+# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.
+
+import numpy as np
+import pylab as plt
+
+import nifty6 as ift
+
+if __name__ == '__main__':
+ dom = ift.UnstructuredDomain(1)
+ a = ift.FieldAdapter(dom, 'a')
+ b = ift.FieldAdapter(dom, 'b')
+
+ uninformative_scaling = 10
+ lh = (a.adjoint @ a).scale(uninformative_scaling) + (b.adjoint @ b).scale(-1.35*2).exp()
+ lh = ift.VariableCovarianceGaussianEnergy(dom, 'a', 'b', np.float64) @ lh
+ icsamp = ift.AbsDeltaEnergyController(deltaE=0.1, iteration_limit=2)
+ icnewton = ift.GradientNormController(iteration_limit=1, name='Mini')
+ ham = ift.StandardHamiltonian(lh, icsamp)
+ newton = ift.SteepestDescent(icnewton)
+ pos = ift.from_random('normal', ham.domain)
+
+ x_limits = [-2/uninformative_scaling, 2/uninformative_scaling]
+ y_limits = [-4, 2]
+ x = np.linspace(*x_limits, num=101)
+ y = np.linspace(*y_limits, num=101)
+ z = np.empty((x.size, y.size))
+ for ii, xx in enumerate(x):
+ for jj, yy in enumerate(y):
+ inp = ift.MultiField.from_raw(lh.domain, {'a': xx, 'b': yy})
+ z[ii, jj] = np.exp(-1*ham(inp).val)
+
+ for ii in range(10):
+ if ii % 2 == 0:
+ mgkl = ift.MetricGaussianKL(pos, ham, 40)
+
+ plt.cla()
+ plt.contour(x*uninformative_scaling, y, z.T)
+ xs, ys = [], []
+ for samp in mgkl.samples:
+ samp = (samp + pos).val
+ xs.append(samp['a'])
+ ys.append(samp['b'])
+ plt.scatter(np.array(xs)*uninformative_scaling, np.array(ys))
+ plt.scatter(pos.val['a'], pos.val['b'], color='red')
+ plt.draw()
+ plt.pause(0.01)
+
+ mgkl, _ = newton(mgkl)
+ pos = mgkl.position
+ ift.logger.info('Finished')
+ # plt.show()
diff --git a/demos/visual_demo.py b/demos/visual_demo.py
deleted file mode 100644
index a60c3bb542e58dcde8f08bd5496024a0fc1d8be8..0000000000000000000000000000000000000000
--- a/demos/visual_demo.py
+++ /dev/null
@@ -1,69 +0,0 @@
-# This program is free software: you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published by
-# the Free Software Foundation, either version 3 of the License, or
-# (at your option) any later version.
-#
-# This program is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with this program. If not, see <http://www.gnu.org/licenses/>.
-#
-# Copyright(C) 2013-2019 Max-Planck-Society
-# Authors: Reimar Leike, Philipp Arras
-#
-# NIFTy is being developed at the Max-Planck-Institut fuer Astrophysik.
-
-import numpy as np
-import pylab as plt
-
-import nifty6 as ift
-
-
-if __name__ == '__main__':
- dom = ift.UnstructuredDomain(1)
- a = ift.FieldAdapter(dom, 'a')
- b = ift.FieldAdapter(dom, 'b')
-
- uninformative_scaling = 10
- ab = (a.adjoint @ a).scale(uninformative_scaling) + (b.adjoint @ b).scale(-1.35*2).exp()
-
- lh = ift.VariableCovarianceGaussianEnergy(dom, 'a', 'b', np.float64) @ ab
- icsamp = ift.AbsDeltaEnergyController(deltaE=0.1,
- iteration_limit=2)
- icnewton = ift.GradientNormController(iteration_limit=1, name='Mini')
- ham = ift.StandardHamiltonian(lh, icsamp)
- newton = ift.SteepestDescent(icnewton)
- # newton = ift.NewtonCG(icnewton, max_cg_iterations=2)
-
- pos = ift.from_random('normal', ham.domain)
- for _ in range(10):
- mgvi = ift.MetricGaussianKL(pos, ham, 40)
- for _ in range(2):
- mgvi, _ = newton(mgvi)
- pos = mgvi.position
-
- plt.cla()
- x_limits = [-2/uninformative_scaling, 2/uninformative_scaling]
- y_limits = [-4, 2]
- x = np.linspace(*x_limits, num=101)
- y = np.linspace(*y_limits, num=101)
- z = np.empty((x.size, y.size))
- for ii, xx in enumerate(x):
- for jj, yy in enumerate(y):
- inp = ift.MultiField.from_raw(lh.domain, {'a': xx, 'b': yy})
- z[ii, jj] = np.exp(-1*ham(inp).val)
- plt.contour(x*uninformative_scaling, y, z.T)
-
- xs, ys = [], []
- for samp in mgvi.samples:
- samp = (samp + pos).val
- xs.append(samp['a'])
- ys.append(samp['b'])
- plt.scatter(np.array(xs)*uninformative_scaling, np.array(ys))
- plt.scatter(pos.val['a'], pos.val['b'], color='red')
-
- plt.draw()
- plt.pause(3/60)