论文

研究用 Concept Induction 解析农业 CNN 中的背景偏差机制

Understanding Contextual Bias in Agricultural CNNs Through Hidden-Neuron Concept Induction

精选理由

训练农业类图像分类模型的人应该看看这篇。它不只测准确率,而是用知识图谱加逻辑表达式解释神经元到底在“看”什么背景特征。27.4% 的验证率和“偏差越强解释性越低”这个反直觉结果,对做模型可解释性和数据集清洗都有直接参考价值。

一项 arXiv 研究考察卷积神经网络如何利用不含目标对象的背景图像块进行分类,并尝试用 Concept Induction 方法揭示隐藏层神经元激活对应的可解释视觉属性。研究覆盖 5 个具有不同背景偏差程度的农业数据集,分析了 4 种 CNN 架构在 3 个网络深度、5 个图像块位置上的表现。

正文

技术上,研究者将每个图像块的定量描述子(含 edge、Gabor、Tamura 纹理特征)编码进知识图谱,作为 Concept Induction 的背景知识,推导出刻画神经元激活的逻辑表达式,再用 Mann-Whitney U 检验和 Cliff's delta 在保留图像上验证。在测试的 48,013 个神经元-特征对中,13,164 个(27.4%)通过验证,且中层验证率(40.2%)远高于深层(12.0%)。

一个反直觉发现是:背景偏差更强的数据集并不意味着神经元-特征验证率更高——中度偏差的 Sugarcane 数据集验证率最高(40.5%),而高度偏差的 Cassava 最低(19.2%)。这说明 CNN 的行为层面偏差依赖与其隐藏神经元的语义可解释性反映的是不同侧面,Concept Induction 可以作为基于图像块偏差检测方法的补充工具。

原文 · arXiv cs.LG

Understanding Contextual Bias in Agricultural CNNs Through Hidden-Neuron Concept Induction

Convolutional Neural Networks (CNNs) can classify images using background-dominated patches that contain little or none of the target object, showing that they exploit contextual information. Previous studies detected this behavior through classification accuracy but say little about what CNN hidden layers represent. We investigate whether Concept Induction can reveal interpretable visual properties associated with hidden-neuron activations on contextual image patches. Using five agricultural datasets with high, moderate, and low contextual bias, we analyze four CNN architectures at three network depths and five patch locations. Quantitative image descriptors of each patch are encoded in a knowledge graph and used as background knowledge for Concept Induction, which derives logical expressions characterizing neuron activation. These expressions are tested on held-out images using the Mann-Whitney U test and Cliff's delta. Of 48,013 tested neuron-feature pairs, 13,164 (27.4%) were validated. Validation was most frequent at intermediate depths (mean 40.2% at mid layers versus 12.0% at deep layers), and no patch location dominated across datasets. Among the tested descriptors, edge, Gabor, and Tamura texture features had the highest validation rates. Stronger contextual bias did not correspond to higher neuron-feature validation rates: moderately biased Sugarcane had the highest rate (40.5%) and highly biased Cassava the lowest (19.2%). These results suggest that behavioral context dependence and the semantic interpretability of hidden neurons capture different aspects of CNN behavior, and that Concept Induction can complement patch-based bias detection.

产业坐标与信源核验

本条情报归属 AITOP「论文」核心追踪栏目。涉及核心实体与技术议题包括:CNN、Concept Induction、可解释性、农业AI,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Abhilekha Dalal, Avishek Das, Samatha Ereshi Akkamahadevi),经 AITOP 多维评分管线结构化提炼。