def score_template_background_section_quantizer(log_template,log_invtemplate,bg,E,return_both_summed=True):
"""
We have a set of features, the size of that set is num_feautures. Then we have a
time parameter
Parameters:
===========
log_template: np.ndarray(ndim=2,dtype=np.float32)
log_template.shape == (template_length,num_features)
log_invtemplate: np.ndarray(ndim=2,dtype=np.float32)
log_invtemplate.shape == (template_length,num_features)
bgd: np.ndarray(ndim=1,dtype=np.float32)
E: np.ndarray(ndim=2,dtype=np.float32)
E.shape == (E_length, num_features)
"""
length = min(E.shape[0],log_template.shape[0])
C_full = log_invtemplate[:length] - log_quantizer(1-bg)
W = log_template[:length] - log_quantizer(bg)
W -= C_full
if return_both_summed:
return (E[:length] * W + C_full).sum()
return (E[:length] * W).sum(), C_full.sum()
def score_templates_background_section_quantizer(log_templates,log_invtemplates,bg,E):
"""
We have a set of features, the size of that set is num_feautures. Then we have a
time parameter which is over axis 0, features are over axis 1
Parameters:
===========
log_template: np.ndarray(ndim=2,dtype=np.float32)
log_template.shape == (template_length,num_features)
log_invtemplate: np.ndarray(ndim=2,dtype=np.float32)
log_invtemplate.shape == (template_length,num_features)
bgd: np.ndarray(ndim=1,dtype=np.float32)
E: np.ndarray(ndim=2,dtype=np.float32)
E.shape == (E_length, num_features)
"""
length = min(E.shape[0],log_template.shape[1])
C_full = np.array(tuple(log_invtemplate[:length] - log_quantizer(1-bg)
for log_invtemplate in log_invtemplates))
W = np.array(tuple(log_template[:length] - log_quantizer(bg)
for log_template in log_templates))
W -= C_full
return max((E[:length] * w + c).sum()
for w,c in zip(W,C_full))