DMD

Dynamic mode decomposition.



In [1]:

    
import sparse_dmd as sd



In [2]:

    
import numpy as np

Speech



In [3]:

    
speech = np.loadtxt('benchmark_signals/speech.txt')



In [4]:

    
snapshots = sd.to_snaps(speech)
dmd = sd.DMD(snapshots)
dmd.compute()



In [5]:

    
dmd.modes









    Out[5]:





array([[ 1.]])



In [6]:

    
dmd.amplitudes









    Out[6]:





array([-74.2772366+0.j])



In [7]:

    
tremor = np.loadtxt('benchmark_signals/tremor.txt')



In [8]:

    
snap_t = sd.to_snaps(tremor)
dmd_t = sd.DMD(snap_t)

# This bit fails for this dataset
#dmd_t.compute()



In [9]:

    
dmd_t









    Out[9]:





<sparse_dmd.sparse_dmd.DMD at 0x11191af10>



In [10]:

    
synthetic = np.loadtxt('benchmark_signals/synthetic.txt')



In [11]:

    
snap_s = sd.to_snaps(synthetic)
dmd_s = sd.DMD(snap_s)



In [12]:

    
dmd_s.compute()



In [13]:

    
dmd_s.modes









    Out[13]:





array([[-1.]])



In [14]:

    
dmd_s.amplitudes









    Out[14]:





array([-0.74198576+0.j])

OK, just realizing that this algo is really for 2D timeseries — so should work on a seismic volume I think. Will try next...



In [ ]: