利用解释的力量进行增量训练:一种基于 LIME 的方法
Harnessing the Power of Explanations for Incremental Training: A LIME-Based Approach
AI总结:
本文提出将 LIME 解释距离引入自定义加权损失,并结合 EWC 进行增量训练,在 Google Speech Commands 关键词识别任务中实现各阶段 0.5%–1.5% 的稳定准确率提升。
摘要:
神经网络预测的可解释性对于理解特征重要性以及获得关于神经网络性能的可解释洞察至关重要。然而,对神经网络结果的解释大多局限于可视化,而寻求将这些解释用作反馈以提升模型性能的工作十分稀少。在本研究中,模型解释被反馈给前馈训练,以帮助模型更好地泛化。为此,本文提出了一种自定义加权损失,其权重通过考虑真实 LIME(Local Interpretable Model-Agnostic Explanations,局部可解释模型无关解释)解释与模型预测的 LIME 解释之间的欧氏距离来生成。此外,在实际训练场景中,由于无法一次性获得所有训练数据,开发一种能够帮助模型按顺序学习而不丢失先前数据分布信息的解决方案势在必行。因此,该框架将自定义加权损失与 Elastic Weight Consolidation(EWC,弹性权重巩固)相结合,以在顺序测试集上保持性能。在使用 Google Speech Commands 数据集进行关键词识别任务时,与基于传统损失的训练方法相比,所提出的自定义训练流程在增量学习设置的所有阶段都使准确率持续提升了 0.5% 至 1.5%。
英文摘要:
Explainability of neural network prediction is essential to understand feature importance and gain interpretable insight into neural network performance. However, explanations of neural network outcomes are mostly limited to visualization, and there is scarce work that looks to use these explanations as feedback to improve model performance. In this work, model explanations are fed back to the feed-forward training to help the model generalize better. To this extent, a custom weighted loss where the weights are generated by considering the Euclidean distances between true LIME (Local Interpretable Model-Agnostic Explanations) explanations and model-predicted LIME explanations is proposed. Also, in practical training scenarios, developing a solution that can help the model learn sequentially without losing information on previous data distribution is imperative due to the unavailability of all the training data at once. Thus, the framework incorporates the custom weighted loss with Elastic Weight Consolidation (EWC) to maintain performance in sequential testing sets. The proposed custom training procedure results in a consistent enhancement of accuracy ranging from 0.5% to 1.5% throughout all phases of the incremental learning setup compared to traditional loss-based training methods for the keyword spotting task using the Google Speech Commands dataset.
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。