Correcting Deep Neural Networks with Small, Generalizing Patches

Correcting Deep Neural Networks with Small, Generalizing Patches
复制标题

使用小型泛化补丁修正深度神经网络

DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Aditya V. Thakur
Aditya V. Thakur
中科院分区：
--
文献类型：
--
作者：
Matthew Sotoudeh;Aditya V. Thakur

文献摘要

被引文献

相似文献

我们考虑了修补深神经网络的问题：为网络权重施加少量更改，以便在网络中产生的CLAS隔离发生了预期的变化。更改网络权重，以便在网络中产生所需的变化，我们使用ACAS XU激励这个问题，这是一个旨在充当飞机避免碰撞的神经网络技术在有限的补丁区域中起作用，基于问题的SMT公式，并且具有许多理想的收敛属性。有效的是，我们引入了神经网络的符号表示，并介绍了Relu神经网络（冷冻网络）。
We consider the problem of patching a deep neural network: applying a small change to the network weights in order to produce a desired change in the clas-sifications made by the network. We motivate this problem using ACAS Xu, a well-studied neural network intended to act as an aircraft collision-avoidance sys-tem. Our technique works over infinite patching regions, is based on an SMT formulation of the problem, and has a number of desirable convergence properties. To make this approach efficient, we introduce a symbolic representation of neural networks and a generalization of ReLU neural networks (Frozen Networks). We show that our approach can produce highly effective and generalizing patches with a very small number of weight changes.