图模型的无限镜检验
The Infinity Mirror Test for Graph Models
摘要:
与其他机器学习模型一样,图模型具有内置的隐式和显式偏差,这通常会以非平凡的方式影响性能。模型的可信度通常通过使用任意数量或组合的图属性将新生成的图与源图进行比较来衡量。因此,生成的图的大小或拓扑结构的差异指示模型中的损失。然而,在许多系统中,编码在损失函数中的错误是微妙的,并且没有被很好地理解。在目前的工作中,我们介绍了无限镜测试分析图模型的鲁棒性。这种简单的压力测试通过反复将模型拟合到自己的输出来工作。假设完美的图模型与源图没有偏差;然而,模型的隐含偏差和假设被无限镜像测试夸大了,暴露了以前被掩盖的潜在问题。通过对成千上万的合成和真实世界的图形实验的分析,我们表明,几个传统的图形模型退化的令人兴奋和信息的方式。我们相信,观察到的退化模式是未来开发更好的图形模型的线索。
英文摘要:
Graph models, like other machine learning models, have implicit and explicit biases built-in, which often impact performance in nontrivial ways. The model's faithfulness is often measured by comparing the newly generated graph against the source graph using any number or combination of graph properties. Differences in the size or topology of the generated graph, therefore, indicate a loss in the model. Yet, in many systems, errors encoded in loss functions are subtle and not well understood. In the present work, we introduce the Infinity Mirror test for analyzing the robustness of graph models. This straightforward stress test works by repeatedly fitting a model to its own outputs. A hypothetically perfect graph model would have no deviation from the source graph; however, the model's implicit biases and assumptions are exaggerated by the Infinity Mirror test, exposing potential issues that were previously obscured. Through an analysis of thousands of experiments on synthetic and real-world graphs, we show that several conventional graph models degenerate in exciting and informative ways. We believe that the observed degenerative patterns are clues to the future development of better graph models.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。