def construct_despot1_cost_function(cost_func, n_images, params, start_range, constraints):
"""
Arguments:
cost_func --
params -- the parameters that are free to vary for each sequence
start_range --
constraints --
"""
DESPOT1_Cost_Function = MultiOptimizationHelper(cost_func, n_images, params=params, start_range=start_range)
indices = DESPOT1_Cost_Function.parameter_indices
# TODO name the constants
for seq in n_images:
# nex must be > 0.01 to avoid singular F
DESPOT1_Cost_Function.add_affine_constraint((seq, 'nex'), 'ineq', 1.0, -0.01)
if 'spgr' in indices and 'theta' in indices['spgr']:
DESPOT1_Cost_Function.add_affine_constraint(('spgr', 'theta'), 'ineq', 1.0, -1./32767.)
# theta must be < 80 for spgr
DESPOT1_Cost_Function.add_affine_constraint(('spgr', 'theta'), 'ineq', -1.0, np.pi*80./180.)
# sort theta ascending
for prev, next in zip(indices['spgr']['theta'][:-1], indices['spgr']['theta'][1:]):
DESPOT1_Cost_Function.constraints.append({
'type': 'ineq',
'fun': lambda x, prev=prev, next=next: x[next] - x[prev]
})
# TODO: modify this to account for differences in tr between seq.
# Fixed time constraint
DESPOT1_Cost_Function.constraints.append({
'type': 'eq',
'fun': lambda x, spgr_nex_idx=indices['spgr']['nex']: sum(x[spgr_nex_idx]) - 10.
})
return DESPOT1_Cost_Function
def construct_diffusion_cost_function(cost_func, n_images, params, start_range, constraints):
"""
Arguments:
cost_func --
params -- the parameters that are free to vary for each sequence
start_range --
constraints --
"""
Cost_Function = MultiOptimizationHelper(cost_func, n_images, params=params, start_range=start_range)
indices = Cost_Function.parameter_indices
# TODO name the constants
# nex must be > 0.01 to avoid singular F
Cost_Function.add_affine_constraint(("diffusion", "nex"), "ineq", 1.0, -0.01)
Cost_Function.add_affine_constraint(("diffusion", "b"), "ineq")
for prev, next in zip(indices["diffusion"]["b"][:-1], indices["diffusion"]["b"][1:]):
Cost_Function.constraints.append({"type": "ineq", "fun": lambda x, prev=prev, next=next: x[next] - x[prev]})
# Fixed time constraint
# TODO: modify this to account for difference in min_tr bet seq.
Cost_Function.constraints.append(
{"type": "eq", "fun": lambda x, nex_idx=indices["diffusion"]["nex"]: sum(x[nex_idx]) - 1.0}
)
return Cost_Function
