arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2106.10185cs.LGcs.AI
收藏

NoiseGrad:通过向模型权重引入随机性来增强简化

NoiseGrad: Enhancing Explanations by Introducing Stochasticity to Model Weights

TU Berlin(柏林工业大学);Understandable Machine Intelligence Lab(可理解机器智能实验室);RIKEN AIP(日本理化学研究所人工智能中心)

Kirill Bykov, Anna Hedström, Shinichi Nakajima, Marina M. -C. Höhne

摘要:

为了揭示黑盒学习机器(如深度神经网络)的决策过程,已经做出了许多努力,从而产生了有用的局部和全局解释方法。对于局部解释,随机性是已知的帮助:一个简单的方法,称为SmoothGrad,通过向输入空间添加噪声并对噪声输入的解释进行平均,提高了基于梯度的归因的视觉质量。在本文中,我们扩展了这一想法,并提出了NoiseGrad,增强了局部和全局的解释方法。具体地,NoiseGrad在权重参数空间中引入随机性,使得决策边界被扰动。由于输入扰动和决策边界扰动之间的双重关系,与SmoothGrad类似,NoiseGrad有望增强局部解释。我们评估NoiseGrad及其融合SmoothGrad - FusionGrad -定性和定量的几个评价标准,并表明我们的新方法显着优于基线方法。NoiseGrad和FusionGrad都是方法不可知的,并且与SmoothGrad一样方便,使用简单的启发式来选择超参数设置,而无需微调。

英文摘要:

Many efforts have been made for revealing the decision-making process of black-box learning machines such as deep neural networks, resulting in useful local and global explanation methods. For local explanation, stochasticity is known to help: a simple method, called SmoothGrad, has improved the visual quality of gradient-based attribution by adding noise to the input space and averaging the explanations of the noisy inputs. In this paper, we extend this idea and propose NoiseGrad that enhances both local and global explanation methods. Specifically, NoiseGrad introduces stochasticity in the weight parameter space, such that the decision boundary is perturbed. NoiseGrad is expected to enhance the local explanation, similarly to SmoothGrad, due to the dual relationship between the input perturbation and the decision boundary perturbation. We evaluate NoiseGrad and its fusion with SmoothGrad -- FusionGrad -- qualitatively and quantitatively with several evaluation criteria, and show that our novel approach significantly outperforms the baseline methods. Both NoiseGrad and FusionGrad are method-agnostic and as handy as SmoothGrad using a simple heuristic for the choice of the hyperparameter setting without the need of finetuning.

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。

↑