Source code for xuance.tensorflow.policies.coordination_graph

import numpy as np
import tensorflow.keras as tk
import tensorflow as tf
import torch
from xuance.tensorflow import Module
try:
    import torch_scatter
except ImportError:
    print("The module torch_scatter is not installed.")


[docs] class DCG_utility(Module): def __init__(self, dim_input, dim_hidden, dim_output): super(DCG_utility, self).__init__() self.dim_input = dim_input self.dim_hidden = dim_hidden self.dim_output = dim_output layers = [tk.layers.Dense(units=self.dim_hidden, activation='relu', input_shape=(self.dim_input,)), tk.layers.Dense(units=self.dim_output, activation=None, input_shape=(self.dim_hidden,))] self.outputs = tk.Sequential(layers) @tf.function def call(self, hidden_states_n, **kwargs): return self.outputs(hidden_states_n)
[docs] class DCG_payoff(DCG_utility): def __init__(self, dim_input, dim_hidden, dim_act, args): self.dim_act = dim_act self.low_rank_payoff = args.low_rank_payoff self.payoff_rank = args.payoff_rank dim_payoff_out = 2 * self.payoff_rank * self.dim_act if self.low_rank_payoff else self.dim_act ** 2 super(DCG_payoff, self).__init__(dim_input, dim_hidden, dim_payoff_out) @tf.function def call(self, hidden_states_n, edges_from=None, edges_to=None, **kwargs): input_payoff_0 = tf.concat([tf.gather(hidden_states_n, edges_from, axis=1), tf.gather(hidden_states_n, edges_to, axis=1)], axis=-1) input_payoff_1 = tf.concat([tf.gather(hidden_states_n, edges_to, axis=1), tf.gather(hidden_states_n, edges_from, axis=1)], axis=-1) input_payoff = tf.stack([input_payoff_0, input_payoff_1], axis=0) input_shape = input_payoff.shape payoffs = self.outputs(tf.reshape(input_payoff, [-1, input_shape[-1]])) payoffs = tf.reshape(payoffs, input_shape[:-1] + (self.dim_output, )) dim = payoffs.shape[0:-1] if self.low_rank_payoff: payoffs = payoffs.view(np.prod(dim) * self.payoff_rank, 2, self.dim_act) payoffs = tf.linalg.matmul(tf.expand_dims(payoffs[:, 0, :], -1), tf.expand_dims(payoffs[:, 1, :], -2)) # (dim_act * 1) * (1 * dim_act) -> (dim_act * dim_act) payoffs = tf.reduce_sum(tf.reshape(payoffs, list(dim) + [self.payoff_rank, self.dim_act, self.dim_act]), axis=-3) else: payoffs = tf.reshape(payoffs, list(dim) + [self.dim_act, self.dim_act]) payoffs = tf.Variable(payoffs) payoffs[1].assign(tf.transpose(payoffs[1], perm=(0, 1, 3, 2))) # f_ij(a_i, a_j) <-> f_ji(a_j, a_i) return tf.reduce_mean(payoffs, axis=0) # f^E_{ij} = (f_ij(a_i, a_j) + f_ji(a_j, a_i)) / 2
[docs] class Coordination_Graph(object): def __init__(self, n_vertexes, graph_type): self.n_vertexes = n_vertexes self.edges = [] if graph_type == "CYCLE": self.edges = [(i, i + 1) for i in range(self.n_vertexes - 1)] + [(self.n_vertexes - 1, 0)] elif graph_type == "LINE": self.edges = [(i, i + 1) for i in range(self.n_vertexes - 1)] elif graph_type == "STAR": self.edges = [(0, i + 1) for i in range(self.n_vertexes - 1)] elif graph_type == "VDN": pass elif graph_type == "FULL": self.edges = [[(j, i + j + 1) for i in range(self.n_vertexes - j - 1)] for j in range(self.n_vertexes - 1)] self.edges = [e for l in self.edges for e in l] else: raise AttributeError("There is no graph type named {}!".format(graph_type)) self.n_edges = len(self.edges) self.edges_from = None self.edges_to = None
[docs] def set_coordination_graph(self): self.edges_from = torch.zeros(self.n_edges).long() self.edges_to = torch.zeros(self.n_edges).long() for i, edge in enumerate(self.edges): self.edges_from[i] = edge[0] self.edges_to[i] = edge[1] self.edges_n_in = torch_scatter.scatter_add(src=self.edges_to.new_ones(len(self.edges_to)), index=self.edges_to, dim=0, dim_size=self.n_vertexes) \ + torch_scatter.scatter_add(src=self.edges_to.new_ones(len(self.edges_to)), index=self.edges_from, dim=0, dim_size=self.n_vertexes) self.edges_n_in = self.edges_n_in.float() return