clean-up, documentation

proxtoolbox/ProxOperators/P_Sparsity.py

@@ -17,10 +17,8 @@ class P_Sparsity(ProxOperator):
         ----------
         config : dict - Dictionary containing the problem configuration. It must contain the following mappings:
         'sparsity_parameter' : int
-        'Ny' : int
         """
         self.sparsity_parameter = config['sparsity_parameter']
-        self.ny = config['Ny']
 
         if 'sparsity_support' in config:
             self.support = config['sparsity_support'].real.astype(np.uint)
@@ -53,13 +51,13 @@ class P_Sparsity(ProxOperator):
         -------
         p_Sparsity : array_like, the projection
         """
-        u *= self.support # apply support
-        p_Sparsity = 0 * u
+        u *= self.support  # apply support (simple 1 if no support)
+        p_sparse = 0 * u
         sorting = np.argsort(abs(u), axis=None)  # gives indices of sorted array in ascending order
         indices = np.asarray([unravel_index(sorting[i], u.shape) for i in range(-1 * self.sparsity_parameter, 0)])
-        p_Sparsity[indices[:, 0], indices[:, 1]] = u[indices[:, 0], indices[:, 1]]
+        p_sparse[indices[:, 0], indices[:, 1]] = u[indices[:, 0], indices[:, 1]]
 
-        return p_Sparsity
+        return p_sparse
 
 
 class P_Sparsity_real(P_Sparsity):

proxtoolbox/ProxOperators/proxoperators.py

@@ -345,7 +345,7 @@ class P_amp(ProxOperator):
         self.amplitude = config['amplitude']
 
     def work(self, u):
-        return magproj(u, self.amplitude) # argument order changed compared to matlab implementation!!!
+        return magproj(u, self.amplitude)  # argument order changed compared to matlab implementation!!!
 
 
 #                      P_SP.m
@@ -386,7 +386,7 @@ class P_SP(ProxOperator):
 # DESCRIPTION:  Projection subroutine for projecting onto Fourier
 #               magnitude constraints
 #
-# INPUT:        Func_params = a data structure with .data = nonegative real FOURIER DOMAIN CONSTRAINT
+# INPUT:        Func_params = a data structure with .data = non-negative real FOURIER DOMAIN CONSTRAINT
 #                             .data_ball is the regularization parameter described in 
 #                             D. R. Luke, Nonlinear Analysis 75 (2012) 1531–1546.
 #                            .TOL2 is an extra tolerance. 
@@ -448,6 +448,11 @@ class Approx_PM_Gaussian(ProxOperator):
 class Approx_PM_Poisson(ProxOperator):
 
     def __init__(self, config):
+        """
+        Test whether guess is close to the measured data, if not, applies magnitude projection to the data
+
+        :param config:
+        """
         self.TOL2 = config['TOL2']
         self.M = config['data']
         self.b = config['data_sq']
@@ -461,15 +466,15 @@ class Approx_PM_Poisson(ProxOperator):
         Ib = self.Ib
         epsilon = self.epsilon
 
-        U = fft2(u)
-        U_sq = U * conj(U)
-        tmp = U_sq / b
-        tmp[Ib] = 1
-        U_sq[Ib] = 0
+        U = fft2(u)  # guess in Fourier space
+        U_sq = U * conj(U)  # modulus squared guess in Fourier space
+        tmp = U_sq / b  # relative error of the guess (1=good)
+        tmp[Ib] = 1  # where data_zeros, set rel.error to 1
+        U_sq[Ib] = 0  # where data_zeros, set guess to zero
         IU = tmp == 0
-        tmp[IU] = 1
+        tmp[IU] = 1  # catch invalid values for the logarithm next line
         tmp = log(tmp)
-        hU = sum(sum(U_sq * tmp + b - U_sq))
+        hU = sum(sum(U_sq * tmp + b - U_sq))  # Kullback-Leibler divergence?
         if hU >= epsilon + TOL2:
             U0 = magproj(U, M)
             return ifft2(U0)