EAGER: Towards robust, interpretable deep learning via communication theory and neuro-inspiration

EAGER：通过沟通理论和神经灵感实现稳健、可解释的深度学习

• 批准号: 2224263
• 负责人: Upamanyu Madhow
• 金额: $ 25万
• 依托单位: University of California-Santa Barbara
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-07-01 至 2025-06-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2224263&HistoricalAwards=false
• 关键词: EAGER; Towards; robust; interpretable; deep

Deep neural networks (DNNs) are attaining great success in an increasing array of applications, yet there remain persistent concerns regarding their lack of interpretability and robustness. The standard approach to training DNNs is to optimize an end-to-end cost function based on variants of gradient descent. This simple approach is flexible, allowing innovation in architectures and applications, and scaling to neural networks with a large number of parameters, given enough data and computational power. Such end-to-end, or top-down training, however, does not provide control over, or understanding of, the features being extracted by the layers of the neural networks. The vulnerability of DNNs to adversarial attacks, for example, is a symptom of this phenomenon. The proposed research seeks to address these drawbacks using ideas from communication theory and neuroscience: the goal is to actively shape the features being extracted by individual layers of the neural network, in addition to training the overall network to attain an end-to-end goal. This research will contribute to curricular enhancements in signal processing and machine learning explored via courses, REU projects and senior capstone projects.The proposed technical approach leverages the existing computational infrastructure for training, while imposing layer-by-layer constraints aimed at producing sparse, strong activations. Drawing on ideas from communication theory, the goal is to learn “matched filters” which enhance the “signal-to-noise ratio (SNR)” at neuron outputs at each layer. One may show that this approach is consistent with Hebbian and anti-Hebbian (HAH) learning as posited in neuroscience, in which neurons that are strongly activated for an input are promoted, with less active neurons being demoted. This work posits enhanced robustness via such an SNR-maximizing strategy, together with additional nonlinear transformations such as divisive normalization borrowed from neuroscience. While preliminary visualizations indicate more interpretable neurons, there is reason to expect sparse, strong activations to be more amenable to quantitative interpretation via statistical and information-theoretic analysis. The goal of the proposed research is two-fold: to gain theoretical insight into HAH-based learning via toy models, and to demonstrate practical gains in robustness and interpretability relative to state of the art DNNs. Experimental evaluations will initially be conducted on image datasets which provide standard performance benchmarks.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

深度神经网络 (DNN) 在越来越多的应用中取得了巨大成功，但人们仍然担心它们缺乏可解释性和鲁棒性。训练 DNN 的标准方法是根据变体优化端到端成本函数。然而，这种简单的方法很灵活，可以在架构和应用程序上进行创新，并且可以扩展到具有大量参数的神经网络，只要有足够的数据和计算能力即可。 ，不提供控制或理解例如，神经网络各层提取的特征就是这种现象的一个症状，该研究旨在利用通信理论和神经科学的思想来解决这些缺陷。除了训练整个网络以实现端到端目标之外，这项研究还将有助于通过课程探索信号处理和机器学习的课程增强， REU 项目和高级顶点所提出的技术方法利用现有的计算基础设施进行训练，同时施加逐层约束，旨在产生稀疏、强的激活，借鉴通信理论的思想，目标是学习“匹配过滤器”，从而增强“每个强层神经元输出的信噪比（SNR）”可能表明这种方法与神经科学中提出的赫布和反赫布（HAH）学习一致，其中神经元针对输入而被激活。这项工作通过这种信噪比最大化策略以及借用神经科学的分裂归一化等额外的非线性变换来增强鲁棒性，虽然初步的可视化更多地表明了可解释的神经元，但有理由期待稀疏。 ，强激活更适合通过统计和信息理论分析进行定量解释。拟议研究的目标有两个：通过玩具模型获得对基于 HAH 的学习的理论见解。展示相对于最先进的 DNN 在鲁棒性和可解释性方面的实际收益。实验评估最初将在提供标准性能基准的图像数据集上进行。该奖项的法定使命是通过使用基金会的智力优势和评估来进行评估，并被认为值得支持。更广泛的影响审查标准。

项目成果

Robust Deep Learning via Layerwise Tilted Exponentials

通过分层倾斜指数进行稳健的深度学习

• 发表时间: 2023-07
• 期刊: ICML 2023
• 作者: Puranik, Bhagyashree;Beirami, Ahmad;Qin, Yao;Madhow, Upamanyu
• 通讯作者: Madhow, Upamanyu

Dynamic Positive Reinforcement For Long-Term Fairness

动态正强化以实现长期公平

• 发表时间: 2022-01
• 期刊: ICML 2022 Workshop on Responsible Decision Making in Dynamic Environments
• 作者: Puranik, B.;Madhow, U.;Pedarsani, R.
• 通讯作者: Pedarsani, R.

Dynamic Positive Reinforcement for Long-Term Fairness

动态正强化以实现长期公平

• 发表时间: 2022-01
• 期刊: ICLR 2022 Workshop on Socially Responsible Machine Learning
• 作者: Puranik, B;Madhow, U.;Pedarsani, R.
• 通讯作者: Pedarsani, R.

Towards robust, interpretable neural networks via Hebbian/anti-Hebbian learning: A software framework for training with feature-based costs

通过赫布/反赫布学习实现稳健、可解释的神经网络：基于特征成本的训练软件框架

• DOI: 10.1016/j.simpa.2022.100347
• 发表时间: 2022-08
• 期刊: Software Impacts
• 作者: Cekic, Metehan;Bakiskan, Can;Madhow, Upamanyu
• 通讯作者: Madhow, Upamanyu

A Dynamic Decision-Making Framework Promoting Long-Term Fairness

促进长期公平的动态决策框架

• DOI: 10.1145/3514094.3534127
• 发表时间: 2022-07
• 期刊: AIES 2022
• 作者: Puranik, Bhagyashree;Madhow, Upamanyu;Pedarsani, Ramtin
• 通讯作者: Pedarsani, Ramtin