Open CloudyDory opened 10 months ago
The following changes are made in bmtk/simulator/filternet/lgnmodel/fitfuns.py to significantly speed up network preparation:
bmtk/simulator/filternet/lgnmodel/fitfuns.py
ff()
np.tile
normalizecols()
makeBasis_StimKernel()
kbasis0
The following changes are made in
bmtk/simulator/filternet/lgnmodel/fitfuns.py
to significantly speed up network preparation:ff()
to use vectorized computation instead of looping over an NumPy array.np.tile
in functionnormalizecols()
andmakeBasis_StimKernel()
by automatic broadcasting in NumPy.makeBasis_StimKernel()
, compute variablekbasis0
in a single step instead of using a for-loop in original code.