def graph_conv_cheby(self, x, W, L, lmax, Fout, K):
# parameters
# B = batch size
# V = nb vertices
# Fin = nb input features
# Fout = nb output features
# K = Chebyshev order & support size
B, V, Fin = x.get_shape()
B, V, Fin = int(B), int(V), int(Fin)
print("fin", Fin)
# rescale Laplacian
lmax = lmaxX(L)
L = rescale_L(L, lmax)
# scipy sparse matric of L
L = L.tocoo()
indices = np.column_stack((L.row, L.col))
L = tf.SparseTensor(indices, L.data, L.shape)
L = tf.sparse_reorder(L)
# Transform to Chebyshev basis
x0 = tf.transpose(x, perm=[1, 2, 0]) # V x Fin x B
x0 = tf.reshape(x0, [V, Fin * B]) # V x Fin*B
x = tf.expand_dims(x0, 0) # 1 x V x Fin*B
def concat(x, x_):
x_ = tf.expand_dims(x_, 0) # 1 x V x Fin*B
return tf.concat([x, x_], 0) # K x V x Fin*B
if K > 1:
x1 = tf.sparse_tensor_dense_matmul(L, x0)
x = concat(x, x1)
for k in range(2, K):
x2 = 2 * tf.sparse_tensor_dense_matmul(L, x1) - x0
x = concat(x, x2) # M x Fin*B
x0, x1 = x1, x2
x = tf.reshape(x, [K, V, Fin, B]) # K x V x Fin x B
x = tf.transpose(x, perm=[3, 1, 2, 0]) # B x V x Fin x K
print("xxxx", x)
x = tf.reshape(x, [B * V, Fin * K]) # B*V x Fin*K
# Compose linearly Fin features to get Fout features
x = tf.matmul(x, W) # B*V x Fout
x = tf.reshape(x, [B, V, Fout]) # B x V x Fout
return x
# Construct graph
t_start = time.time()
grid_side = 28
number_edges = 8
metric = 'euclidean'
A = grid_graph(grid_side,number_edges,metric) # create graph of Euclidean grid
# Compute coarsened graphs
coarsening_levels = 4
num_vertices, L, perm = coarsen(A, coarsening_levels)
# Compute max eigenvalue of graph Laplacians
lmax = []
for i in range(coarsening_levels):
lmax.append(lmaxX(L[i]))
print('lmax: ' + str([lmax[i] for i in range(coarsening_levels)]))
train_data = tf.Session().run(train_data)
#comment this line for non rotated test images
test_data = tf.Session().run(test_data)
# Reindex nodes to satisfy a binary tree structure
train_data = perm_data(train_data, perm)
val_data = perm_data(val_data, perm)
test_data = perm_data(test_data, perm)
print(train_data.shape)
print(val_data.shape)
print(test_data.shape)