def _receptive_fn(self, nodes):
if nodes.type == 'u':
neighs_num = self.u_neighs_num
return self.graph.V(nodes.type, nodes).alias('v').outV("u-i")\
.sample(neighs_num[0]).by("random").alias('v1')\
.emit(lambda x: gl.EgoGraph(x['v'], [gl.Layer(nodes=x['v1'])]))
else:
neighs_num = self.i_neighs_num
return self.graph.V(nodes.type, nodes).alias('v').inV("u-i")\
.sample(neighs_num[0]).by("random").alias('v1')\
.emit(lambda x: gl.EgoGraph(x['v'], [gl.Layer(nodes=x['v1'])]))
def _receptive_fn(self, nodes):
alias = ['v' + str(i + 1) for i in range(self.hops_num)]
# sample based
if self.neighs_num is not None:
assert len(self.neighs_num) == self.hops_num
sample_func = lambda v, params: v.outV(self.edge_type).sample(params).by('topk')
return self.graph.V(nodes.type, feed=nodes).alias('v')\
.repeat(sample_func, self.hops_num, params_list=self.neighs_num, alias_list=alias)\
.emit(lambda x: gl.EgoGraph(x['v'], [gl.Layer(nodes=x[name]) for name in alias]))
# full batch based
sample_func = lambda v: v.outV(self.edge_type).sample().by('full')
return self.graph.V(nodes.type, feed=nodes).alias('v')\
.repeat(sample_func, self.hops_num, alias_list=alias)\
.emit(lambda x: gl.EgoGraph(x['v'], [gl.Layer(nodes=x[name]) for name in alias]))