稀疏概念瓶颈模型：对比学习中的 Gumbel 技巧
Sparse Concept Bottleneck Models: Gumbel Tricks in Contrastive Learning
AI总结：
本文提出 Sparse-CBM，利用概念瓶颈层、Gumbel-Softmax 稀疏损失与 Concept Matrix Search，从类 CLIP 模型构建可解释分类器并提升复杂数据集准确率。
摘要：
我们提出了一种使用概念瓶颈模型（Concept Bottleneck Models，CBM）进行可解释分类的新型架构和方法。尽管图像分类任务中的 SOTA 方法像黑箱一样工作，但人们对能够提供可解释结果的模型的需求正在不断增长。这类模型通常利用目标实例的额外描述（称为概念）来学习预测类别标签上的分布。然而，现有瓶颈方法存在若干局限：其准确率低于标准模型，并且 CBM 需要额外的概念集合才能发挥作用。我们提供了一个框架，可由预训练多模态编码器和新的类 CLIP 架构创建概念瓶颈模型。通过引入一种称为概念瓶颈层的新型层，我们概述了训练这些层的三种方法：使用 ℓ1 损失、对比损失以及基于 Gumbel-Softmax 分布的损失函数（Sparse-CBM），而最后的全连接层仍使用交叉熵进行训练。我们表明，在基于 CLIP 的瓶颈模型中使用稀疏隐藏层可显著提升准确率。这意味着概念激活向量的稀疏表示在概念瓶颈模型中是有意义的。此外，借助我们的 Concept Matrix Search 算法，无需任何额外训练或微调即可改善 CLIP 在复杂数据集上的预测。代码可在 https://github.com/Andron00e/SparseCBM 获取。
英文摘要：
We propose a novel architecture and method of explainable classification with Concept Bottleneck Models (CBMs). While SOTA approaches to Image Classification task work as a black box, there is a growing demand for models that would provide interpreted results. Such a models often learn to predict the distribution over class labels using additional description of this target instances, called concepts. However, existing Bottleneck methods have a number of limitations: their accuracy is lower than that of a standard model and CBMs require an additional set of concepts to leverage. We provide a framework for creating Concept Bottleneck Model from pre-trained multi-modal encoder and new CLIP-like architectures. By introducing a new type of layers known as Concept Bottleneck Layers, we outline three methods for training them: with $\ell_1$-loss, contrastive loss and loss function based on Gumbel-Softmax distribution (Sparse-CBM), while final FC layer is still trained with Cross-Entropy. We show a significant increase in accuracy using sparse hidden layers in CLIP-based bottleneck models. Which means that sparse representation of concepts activation vector is meaningful in Concept Bottleneck Models. Moreover, with our Concept Matrix Search algorithm we can improve CLIP predictions on complex datasets without any additional training or fine-tuning. The code is available at: https://github.com/Andron00e/SparseCBM.
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。
