arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 1909.11275cs.LGstat.ML
收藏

用于神经网络可解释性的切换线性投影

Switched linear projections for neural network interpretability

University of Otago(奥塔哥大学)

Lech Szymanski, Brendan McCane, Craig Atkinson

摘要:

我们引入切换线性投影,用于根据输入空间中的单个线性投影来表达深度神经网络中神经元的活动。该方法的工作原理是隔离活动子网络,一系列线性变换,确定了给定输入实例的网络的整个计算。通过这些投影,我们可以将任何隐藏层中的活动分解为在给定输入实例中检测到的模式。我们还提出,在ReLU网络中,检查使隐藏层中的神经元失活的模式是有指导意义和有意义的,而现有的可解释性方法只跟踪网络计算的活跃方面,这是隐含忽略的。

英文摘要:

We introduce switched linear projections for expressing the activity of a neuron in a deep neural network in terms of a single linear projection in the input space. The method works by isolating the active subnetwork, a series of linear transformations, that determine the entire computation of the network for a given input instance. With these projections we can decompose activity in any hidden layer into patterns detected in a given input instance. We also propose that in ReLU networks it is instructive and meaningful to examine patterns that deactivate the neurons in a hidden layer, something that is implicitly ignored by the existing interpretability methods tracking solely the active aspect of the network's computation.

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。

↑