def TGCN(_X, weights, biases):
###
cell_1 = tgcnCell(gru_units, adj, num_nodes=num_nodes)
cell = tf.nn.rnn_cell.MultiRNNCell([cell_1], state_is_tuple=True)
_X = tf.unstack(_X, axis=1)
# print(_X)
outputs, states = tf.nn.static_rnn(cell, _X, dtype=tf.float32)
# print('000',outputs)
# out = outputs[-1]
# out = tf.reshape(outputs,shape=[seq_len, -1, num_nodes,gru_units])
out = tf.concat(outputs, axis=0)
out = tf.reshape(out, shape=[seq_len,-1,num_nodes,gru_units])
out = tf.transpose(out, perm=[1,0,2,3])
# out = tf.reshape(out, shape=[-1, num_nodes, gru_units])
# print('111',out)
# out = tf.transpose(out,perm=[1,2,0,3]) #n,t,nodes*u
# out = tf.transpose(out,perm=[1,0,2]) #n,t,nodes*u
# print('111',out)
# out = tf.reshape(out,[-1,seq_len,gru_units])
# lstm_last_output = m[-1]
last_output,alpha = self_attention1(out, weight_att, bias_att)
# print('yyy')
# output = tf.reshape(last_output,shape=[-1,num_nodes,gru_units])
# output = tf.reshape(last_output,shape=[-1,gru_units])
output = tf.reshape(last_output,shape=[-1,seq_len])
output = tf.matmul(output, weights['out']) + biases['out']
output = tf.reshape(output,shape=[-1,num_nodes,pre_len])
output = tf.transpose(output, perm=[0,2,1])
output = tf.reshape(output, shape=[-1,num_nodes])
return output, outputs, states, alpha
def TGCN(_X, _weights, _biases):
###
cell_1 = tgcnCell(gru_units,
adj,
num_nodes=num_nodes,
num_features=num_features)
cell = tf.nn.rnn_cell.MultiRNNCell([cell_1], state_is_tuple=True)
print("------------TCGN--------------")
print("_X.shape: ", _X.shape)
_X = tf.unstack(_X, axis=1)
print("After, len(_X): ", len(_X))
print("_X[0].shape: ", _X[0].shape)
outputs, states = tf.nn.static_rnn(cell, _X, dtype=tf.float32)
m = []
print("---------------before len(outputs)----------------")
print(len(outputs))
print(type(outputs[0]))
print(outputs[0])
for i in outputs:
o = tf.reshape(i, shape=[-1, num_nodes, gru_units])
# comment the line below makes no difference
o = tf.reshape(o, shape=[-1, gru_units])
m.append(o)
last_output = m[-1]
print("last_output.shape", last_output.shape)
last_output = last_output[:, 0]
print("last_output.shape", last_output.shape)
# outputs = gru_units * number of nodes in gcn
output = tf.matmul(last_output, _weights['out']) + _biases['out']
# num_nodes * pre_len, 3 in this case
output = tf.reshape(output, shape=[-1, num_nodes, pre_len])
output = tf.transpose(output, perm=[0, 2, 1])
output = tf.reshape(output, shape=[-1, num_nodes])
# pre_len * num_nodes
return output, m, states
def TGCN(_X, _weights, _biases):
###
cell_1 = tgcnCell(gru_units, adj, num_nodes=num_nodes)
cell = tf.nn.rnn_cell.MultiRNNCell([cell_1], state_is_tuple=True)
_X = tf.unstack(_X, axis=1)
outputs, states = tf.nn.static_rnn(cell, _X, dtype=tf.float32)
m = []
for i in outputs:
o = tf.reshape(i, shape=[-1, num_nodes, gru_units])
o = tf.reshape(o, shape=[-1, gru_units])
m.append(o)
last_output = m[-1]
output = tf.matmul(last_output, _weights['out']) + _biases['out']
output = tf.reshape(output, shape=[-1, num_nodes, pre_len])
output = tf.transpose(output, perm=[0, 2, 1])
output = tf.reshape(output, shape=[-1, num_nodes])
return output, m, states
def TGCN_att(_X, weights, biases, tmp_adj, keep_rate):
###
tmp_adj = tf.sparse.from_dense(tmp_adj)
cell_1 = tgcnCell(gru_units, tmp_adj, keep_rate, num_nodes=num_nodes)
cell = tf.nn.rnn_cell.MultiRNNCell([cell_1], state_is_tuple=True)
_X = tf.unstack(_X, axis=1)
outputs, states = tf.nn.static_rnn(cell, _X, dtype=tf.float32)
out = tf.concat(outputs, axis=0)
out = tf.reshape(out, shape=[seq_len, -1, num_nodes, gru_units])
out = tf.transpose(out, perm=[1, 0, 2, 3])
last_output, alpha = self_attention1(out, weight_att, bias_att)
output = tf.reshape(last_output, shape=[-1, seq_len])
output = tf.matmul(output, weights['out']) + biases['out']
output = tf.reshape(output, shape=[-1, num_nodes, pre_len])
output = tf.transpose(output, perm=[0, 2, 1])
output = tf.reshape(output, shape=[-1, num_nodes])
return output, outputs, states
