Selective harvesting over networks

Abstract

Active search on graphs focuses on collecting certain labeled nodes (targets) given global knowledge of the network topology and its edge weights (encoding pairwise similarities) under a query budget constraint. However, in most current networks, nodes, network topology, network size, and edge weights are all initially unknown. In this work we introduce selective harvesting, a variant of active search where the next node to be queried must be chosen among the neighbors of the current queried node set; the available training data for deciding which node to query is restricted to the subgraph induced by the queried set (and their node attributes) and their neighbors (without any node or edge attributes). Therefore, selective harvesting is a sequential decision problem, where we must decide which node to query at each step. A classifier trained in this scenario can suffer from what we call a tunnel vision effect: without any recourse to independent sampling, the urge to only query promising nodes forces classifiers to gather increasingly biased training data, which we show significantly hurts the performance of active search methods and standard classifiers. We demonstrate that it is possible to collect a much larger set of targets by using multiple classifiers, not by combining their predictions as a weighted ensemble, but switching between classifiers used at each step, as a way to ease the tunnel vision effect. We discover that switching classifiers collects more targets by (a) diversifying the training data and (b) broadening the choices of nodes that can be queried in the future. This highlights an exploration, exploitation, and diversification trade-off in our problem that goes beyond the exploration and exploitation duality found in classic sequential decision problems. Based on these observations we propose D 3 TS, a method based on multi-armed bandits for non-stationary stochastic processes that enforces classifier diversity, which outperforms all competing methods on five real network datasets in our evaluation and exhibits comparable performance on the other two.

Active Search of Connections for Case Building and Combating Human Trafficking

“…Our proposed RedThread solution for Active Search of Connections, also queries a limited number of labels from the user but to locally retrieve the maximum number of nodes related and relevant to the given seed. This is closer to active exploration [15,28] (Fig. 1c) which also finds connected and relevant nodes.…”

“…RedThread however is proposing a semi-supervised or active paradigm, where we are using the labels on the nodes, provided by the expert user, as the true clustering. In this sense, RedThread is also closely related to active exploration [15] or selective harvesting [15,28]. These methods however assume the graph is (partially) observable, which is not true in our case.…”

Desenvolvimento de um Algoritmo de Busca por Vértices Específicos em Redes

“…O problema de busca em redes, exatamente como foi apresentado na Seção 1 (com descoberta gradual da rede e utilizando informação dos vértices), foi formulado e tratado recentemente em [Murai et al 2017]. Empregando e adaptando métodos desenvolvidos para outros problemas ( [Avrachenkov et al 2014], [Wang et al 2013], [Pfeiffer et al 2012] e [Bnaya et al 2013]), além de classificadores clássicos de Aprendizado de Máquina, os autores chegaram a conclusão de queé possível obter uma eficiência muito maior variando o classificador (ou método) utilizado em cada passo da busca.…”

“…Entretanto, alguns métodos propostos para Active Search podem ser adaptados para o problema em questão. Por exemplo, o método para Active Search proposto em [Wang et al 2013] foi adaptado em [Murai et al 2017] para considerar apenas a rede conhecida até o momento presente da busca.…”

“…Desta forma, um processo eficiente de mineração dos vértices deve maximinar o número de vértices T ao explorar a rede. O problema de minerar uma rede em busca de vértices com determinadas características encontra aplicações em muitos cenários, tais como campanhas de marketing personalizadas, sistemas de recomendação, detecção de fraudes de segurança, ou mesmo na identificação de estudantes que fumam para uma campanha antitabagismo eficaz [Murai et al 2017]. Resolver este problema de forma eficienteé fundamental: considere o problema de encontrar pessoas no Facebook que declaram morar em Búzios.…”