通过最大张力图像理解抑制机制
Understanding Inhibition Through Maximally Tense Images
AI总结:
本文提出通过“最大张力图像”研究视觉模型中的特征抑制机制,并引入正负归因反演和归因图谱两种可视化技术,同时分析了叠加现象对归因解释造成的干扰。
摘要:
我们研究视觉模型中“特征抑制”的功能作用;即神经网络通过何种机制确保图像不表达某一给定特征?我们观察到,由于ReLU激活函数引入的不对称性,文献中的标准可解释性工具并不直接适用于抑制情形。鉴于此,我们提出应通过研究“最大张力图像”(MTIs)来理解抑制,即那些同时激发和抑制某一给定特征的图像。我们展示了如何利用两种新颖的可视化技术来研究MTIs:正负归因反演(+/- attribution inversions),它将单张图像分解为激发分量和抑制分量;以及归因图谱(attribution atlas),它提供了图像激发/抑制某一特征的各种方式的全局可视化。最后,我们探讨了叠加(superposition)带来的困难,因为这类干扰特征会诱导出与MTIs相同的归因模式。
英文摘要:
We address the functional role of 'feature inhibition' in vision models; that is, what are the mechanisms by which a neural network ensures images do not express a given feature? We observe that standard interpretability tools in the literature are not immediately suited to the inhibitory case, given the asymmetry introduced by the ReLU activation function. Given this, we propose inhibition be understood through a study of 'maximally tense images' (MTIs), i.e. those images that excite and inhibit a given feature simultaneously. We show how MTIs can be studied with two novel visualization techniques; +/- attribution inversions, which split single images into excitatory and inhibitory components, and the attribution atlas, which provides a global visualization of the various ways images can excite/inhibit a feature. Finally, we explore the difficulties introduced by superposition, as such interfering features induce the same attribution motif as MTIs.
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。