config.py

import numpy as np

class RepMethod():
	def __init__(self, 
				align_info = None, 
				p=None, 
				k=10, 
				max_layer=None, 
				alpha = 0.1, 
				num_buckets = None, 
				normalize = True, 
				gammastruc = 1, 
				gammaattr = 1):
		self.p = p #sample p points
		self.k = k #control sample size
		self.max_layer = max_layer #furthest hop distance up to which to compare neighbors
		self.alpha = alpha #discount factor for higher layers
		self.num_buckets = num_buckets #number of buckets to split node feature values into #CURRENTLY BASE OF LOG SCALE
		self.normalize = normalize #whether to normalize node embeddings
		self.gammastruc = gammastruc #parameter weighing structural similarity in node identity
		self.gammaattr = gammaattr #parameter weighing attribute similarity in node identity

class Graph():
	#Undirected, unweighted
	def __init__(self, 
				adj, 
				num_buckets=None, 
				node_labels = None, 
				edge_labels = None,
				graph_label = None, 
				node_attributes = None, 
				true_alignments = None):
		self.G_adj = adj #adjacency matrix
		self.N = self.G_adj.shape[0] #number of nodes
		self.node_degrees = np.ravel(np.sum(self.G_adj, axis=0).astype(int))
		self.max_degree = max(self.node_degrees)
		self.num_buckets = num_buckets #how many buckets to break node features into

		self.node_labels = node_labels
		self.edge_labels = edge_labels
		self.graph_label = graph_label
		self.node_attributes = node_attributes #N x A matrix, where N is # of nodes, and A is # of attributes
		self.kneighbors = None #dict of k-hop neighbors for each node
		self.true_alignments = true_alignments #dict of true alignments, if this graph is a combination of multiple graphs