add new parameters to DMDBase subclasses

pydmd/cdmd.py

@@ -48,14 +48,18 @@ class CDMD(DMDBase):
         numpy.ndarray
     :param bool opt: flag to compute optimal amplitudes. See :class:`DMDBase`.
         Default is False.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
     """
 
     def __init__(self,
                  svd_rank=0,
                  tlsq_rank=0,
                  compression_matrix='uniform',
-                 opt=False):
-        super(CDMD, self).__init__(svd_rank, tlsq_rank, True, opt)
+                 opt=False,
+                 rescale_mode=None):
+        super(CDMD, self).__init__(svd_rank, tlsq_rank, True, opt, rescale_mode)
         self.compression_matrix = compression_matrix
 
     def _compress_snapshots(self):

pydmd/dmd.py

@@ -30,6 +30,9 @@ class DMD(DMDBase):
         Default is False.
     :param bool opt: flag to compute optimal amplitudes. See :class:`DMDBase`.
         Default is False.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
     """
 
     def fit(self, X):

pydmd/dmdc.py

@@ -32,14 +32,19 @@ class DMDc(DMDBase):
         for the `_fit_B_unknown` method of this class. It should be greater or
         equal than `svd_rank`. For the possible values please refer to the
         `svd_rank` parameter description above.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
     :type svd_rank_omega: int or float
     """
-    def __init__(self, svd_rank=0, tlsq_rank=0, opt=False, svd_rank_omega=-1):
+    def __init__(self, svd_rank=0, tlsq_rank=0, opt=False, svd_rank_omega=-1,
+        rescale_mode=None):
         self.svd_rank = svd_rank
         self.tlsq_rank = tlsq_rank
         self.opt = opt
         self.svd_rank_omega = svd_rank_omega
         self.original_time = None
+        self.rescale_mode = rescale_mode
 
         self._eigs = None
         self._Atilde = None
@@ -142,7 +147,7 @@ def _fit_B_unknown(self, X, Y):
         omega = np.vstack([X, self._controlin])
 
         Up, sp, Vp = self._compute_svd(omega, self.svd_rank_omega)
-        
+
         Up1 = Up[:self._snapshots.shape[0], :]
         Up2 = Up[self._snapshots.shape[0]:, :]
 

pydmd/fbdmd.py

@@ -24,6 +24,9 @@ class FbDMD(DMDBase):
         Default is False.
     :param bool opt: flag to compute optimal amplitudes. See :class:`DMDBase`.
         Default is False.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
 
     Reference: Dawson et al. https://arxiv.org/abs/1507.02264
     """

pydmd/hodmd.py

@@ -24,12 +24,17 @@ class HODMD(DMDBase):
         Default is False.
     :param bool opt: flag to compute optimal amplitudes. See :class:`DMDBase`.
         Default is False.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
     :param int d: the new order for spatial dimension of the input snapshots.
         Default is 1.
     """
 
-    def __init__(self, svd_rank=0, tlsq_rank=0, exact=False, opt=False, d=1):
-        super(HODMD, self).__init__(svd_rank, tlsq_rank, exact, opt)
+    def __init__(self, svd_rank=0, tlsq_rank=0, exact=False, opt=False, d=1,
+        rescale_mode=None):
+        super(HODMD, self).__init__(svd_rank, tlsq_rank, exact, opt,
+            rescale_mode)
         self.d = d
 
     def fit(self, X):

pydmd/mrdmd.py

@@ -32,6 +32,9 @@ class MrDMD(DMDBase):
         Default is False.
     :param bool opt: flag to compute optimal amplitudes. See :class:`DMDBase`.
         Default is False.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
     :param int max_cycles: the maximum number of mode oscillations in any given
         time scale. Default is 1.
     :param int max_level: the maximum number of levels. Defualt is 6.
@@ -43,8 +46,10 @@ def __init__(self,
                  exact=False,
                  opt=False,
                  max_cycles=1,
-                 max_level=6):
-        super(MrDMD, self).__init__(svd_rank, tlsq_rank, exact, opt)
+                 max_level=6,
+                 rescale_mode=None):
+        super(MrDMD, self).__init__(svd_rank, tlsq_rank, exact, opt,
+            rescale_mode)
         self.max_cycles = max_cycles
         self.max_level = max_level
         self._nsamples = None

pydmd/optdmd.py

@@ -18,7 +18,7 @@ def pinv_diag(x):
     Utility function to compute the pseudo-inverse of a diagonal matrix.
 
     :param array_like x: diagonal of the matrix to be pseudo-inversed.
-    :return: the computed pseudo-inverse 
+    :return: the computed pseudo-inverse
     :rtype: numpy.ndarray
     """
     t = x.dtype.char.lower()
@@ -59,17 +59,21 @@ class OptDMD(DMDBase):
         Default is False.
     :param bool opt: flag to compute optimal amplitudes. See :class:`DMDBase`.
         Default is False.
+    :param numpy.array rescale_mode: None means no rescaling, empty array means
+        automatic rescaling using SV, otherwise the user chooses the preferred
+        scaling.
     """
 
     def __init__(self,
                  factorization="evd",
                  svd_rank=-1,
                  tlsq_rank=0,
                  exact=False,
-                 opt=False
-                 ):
+                 opt=False,
+                 rescale_mode=None):
 
-        super(OptDMD, self).__init__(svd_rank, tlsq_rank, exact, opt)
+        super(OptDMD, self).__init__(svd_rank, tlsq_rank, exact, opt,
+            rescale_mode)
         self.factorization = factorization
 
         self._svds = None
@@ -129,7 +133,7 @@ def fit(self, X, Y=None):
             _, self._input_space, self._output_space = self._eig_from_lowrank_op(self._Atilde, Uz, Q)
 
         self._modes = self._output_space
-        
+
         return self
 
     def predict(self, X):