fixed test

4 files changed, 35 insertions, 19 deletions

diff --git a/Wrappers/Python/ccpi/framework/VectorData.py b/Wrappers/Python/ccpi/framework/VectorData.py
index fdce3a5..d0fed10 100755
--- a/Wrappers/Python/ccpi/framework/VectorData.py
+++ b/Wrappers/Python/ccpi/framework/VectorData.py
@@ -54,5 +54,5 @@ class VectorData(DataContainer):
                     out = array
                 else:
                     raise ValueError('Incompatible size: expecting {} got {}'.format((self.length,), array.shape))
-        deep_copy = False
+        deep_copy = True
         super(VectorData, self).__init__(out, deep_copy, None)
diff --git a/Wrappers/Python/ccpi/framework/framework.py b/Wrappers/Python/ccpi/framework/framework.py
index b972be6..f91343c 100755
--- a/Wrappers/Python/ccpi/framework/framework.py
+++ b/Wrappers/Python/ccpi/framework/framework.py
@@ -329,7 +329,7 @@ class DataContainer(object):
         self.dimension_labels = {}
         self.geometry = None # Only relevant for AcquisitionData and ImageData
         
-        if dimension_labels is not None and \
+        if dimension_labels != {} and \
            len (dimension_labels) == self.number_of_dimensions:
             for i in range(self.number_of_dimensions):
                 self.dimension_labels[i] = dimension_labels[i]
@@ -632,6 +632,7 @@ class DataContainer(object):
             
     def pixel_wise_binary(self, pwop, x2, *args,  **kwargs):    
         out = kwargs.get('out', None)
+        
         if out is None:
             if isinstance(x2, (int, float, complex)):
                 out = pwop(self.as_array() , x2 , *args, **kwargs )
diff --git a/Wrappers/Python/ccpi/optimisation/algorithms/FISTA.py b/Wrappers/Python/ccpi/optimisation/algorithms/FISTA.py
index db4e8b7..4754d9c 100755
--- a/Wrappers/Python/ccpi/optimisation/algorithms/FISTA.py
+++ b/Wrappers/Python/ccpi/optimisation/algorithms/FISTA.py
@@ -62,28 +62,18 @@ class FISTA(Algorithm):
     def update(self):
 
         self.f.gradient(self.y, out=self.u)
-        print ('update, self.u' , self.u.as_array())
         self.u.__imul__( -self.invL )
         self.u.__iadd__( self.y )
-        print ('update, self.u' , self.u.as_array())
-        
         # x = g.prox(u,invL)
-        print ('update, self.x pre prox' , self.x.as_array())
         self.g.proximal(self.u, self.invL, out=self.x)
-        print ('update, self.x post prox' , self.x.as_array())
         
         self.t = 0.5*(1 + numpy.sqrt(1 + 4*(self.t_old**2)))
         
         self.x.subtract(self.x_old, out=self.y)
-        print ('update, self.y' , self.y.as_array())
-        
         self.y.__imul__ ((self.t_old-1)/self.t)
-        print ('update, self.x' , self.x.as_array())
         self.y.__iadd__( self.x )
-        print ('update, self.y' , self.y.as_array())
         
         self.x_old.fill(self.x)
-        print ('update, self.x_old' , self.x_old.as_array())
         self.t_old = self.t            
         
     def update_objective(self):
diff --git a/Wrappers/Python/wip/fix_test.py b/Wrappers/Python/wip/fix_test.py
index 7b19910..094f571 100755
--- a/Wrappers/Python/wip/fix_test.py
+++ b/Wrappers/Python/wip/fix_test.py
@@ -60,11 +60,11 @@ class Norm1(Function):
 
 opt = {'memopt': True}
 # Problem data.
-m = 3
-n = 3
+m = 30
+n = 30
 np.random.seed(1)
-#Amat = np.asarray( np.random.randn(m, n), dtype=numpy.float32)
-Amat = np.asarray(np.eye(m), dtype=np.float32) * 2
+Amat = np.asarray( np.random.randn(m, n), dtype=numpy.float32)
+#Amat = np.asarray(np.eye(m), dtype=np.float32) * 2
 A = LinearOperatorMatrix(Amat)
 bmat = np.asarray( np.random.randn(m), dtype=numpy.float32)
 bmat *= 0 
@@ -92,8 +92,15 @@ g0 = ZeroFunction()
 #f.L = 30.003
 x_init = vgx.allocate(VectorGeometry.RANDOM, dtype=numpy.float32)
 
+a = VectorData(x_init.as_array(), deep_copy=True)
+
+assert id(x_init.as_array()) != id(a.as_array())
+
+#%%
 f.L = LinearOperator.PowerMethod(A, 25, x_init)[0] * 2
 print ('f.L', f.L)
+rate = (1 / f.L) / 3
+f.L *= 6
 
 # Initial guess
 #x_init = DataContainer(np.zeros((n, 1)))
@@ -102,6 +109,7 @@ print ('b', b.as_array())
 # Create 1-norm object instance
 g1_new = lam * L1Norm()
 g1 = Norm1(lam)
+
 g1 = ZeroFunction()
 #g1(x_init) 
 x = g1.prox(x_init, 1/f.L ) 
@@ -112,6 +120,16 @@ print ("g1.proximal ", x.as_array())
 x = g1_new.proximal(x_init, 0.03 ) 
 print ("g1.proximal ", x.as_array())
 
+x1 = vgx.allocate(VectorGeometry.RANDOM, dtype=numpy.float32)
+pippo = vgx.allocate()
+
+print ("x_init", x_init.as_array())
+print ("x1", x1.as_array())
+a = x_init.subtract(x1, out=pippo)
+
+print ("pippo", pippo.as_array())
+print ("x_init", x_init.as_array())
+print ("x1", x1.as_array())
 
 
 # Combine with least squares and solve using generic FISTA implementation
@@ -120,8 +138,14 @@ def callback(it,  objective, solution):
     print (it, objective, solution.as_array())
 
 fa = FISTA(x_init=x_init, f=f, g=g1)
-fa.max_iteration = 10
-fa.run(3, callback = callback, verbose=False)
+fa.max_iteration = 1000
+fa.update_objective_interval = int( fa.max_iteration / 10 ) 
+fa.run(fa.max_iteration, callback = None, verbose=True)
+
+gd = GradientDescent(x_init=x_init, objective_function=f, rate = rate )
+gd.max_iteration = 100000
+gd.update_objective_interval = 10000
+gd.run(gd.max_iteration, callback = None, verbose=True)
 
 
 # Print for comparison
@@ -130,4 +154,5 @@ print(fa.get_output().as_array())
 print(fa.get_last_objective())
 
 print (A.direct(fa.get_output()).as_array())
-print (b.as_array())
\ No newline at end of file
+print (b.as_array())
+print (A.direct(gd.get_output()).as_array())