def vertex_weights(self):
## first need to pick a simplex to sample from ##
vols = []
n_verts = []
for ii in range(self.n_simplex):
vols.append(complex_volume(self, ii))
n_verts.append(len(self.simplicial_complex[ii]))
norm = sum(vols)
# print(norm.item())
## error catch for 0 volume simplexes
if norm == 0:
simp_ind = np.random.randint(self.n_simplex)
else:
vol_cumsum = np.cumsum([vv/norm for vv in vols])
# print(vol_cumsum)
simp_ind = np.min(np.where(np.random.rand(1) < vol_cumsum)[0])
## sample weights for simplex
exps = [-(torch.rand(1)).log().item() for _ in range(n_verts[simp_ind])]
total = sum(exps)
exps = [exp/total for exp in exps]
## now assign vertex weights out
vert_weights = [0] * self.n_vert
for ii, vert in enumerate(self.simplicial_complex[simp_ind]):
vert_weights[vert] = exps[ii]
return vert_weights
def forward(self, complex_modex):
## first need to pick a simplex to sample from ##
vols = []
n_verts = []
for ii in range(self.n_simplex):
vols.append(complex_volume(complex_model, ii))
n_verts.append(len(complex_model.simplexes[ii]))
norm = sum(vols)
vol_cumsum = np.cumsum([vv/norm for vv in vols])
simp_ind = np.min(np.where(np.random.rand(1) < vol_cumsum)[0])
## sample weights for simplex
exps = [-(torch.rand(1)).log().item() for _ in range(n_verts[simp_ind])]
total = sum(exps)
exps = [exp/total for exp in exps]
## now assign vertex weights out
vert_weights = [0] * complex_model.n_vert
for ii, vert in enumerate(complex_model.simplexes[simp_ind]):
vert_weights[vert] = exps[ii]
return vert_weights
def total_volume(self, vol_function=complex_volume):
vol = 0
# for simp in range(self.n_simplex):
# vol += complex_volume(self, simp)
vol = complex_volume(self, 0)
return vol