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 1803.08337cs.CVcs.LG
收藏

深度网络喜欢看什么?

What do Deep Networks Like to See?

German Research Center for Artificial Intelligence (DFKI)(德国人工智能研究中心);TU Kaiserslautern(凯撒斯劳滕工业大学)

Sebastian Palacio, Joachim Folz, Jörn Hees, Federico Raue, Damian Borth, Andreas Dengel

摘要:

我们提出了一种新的方法来测量和理解卷积神经网络,通过量化他们让输入信号的数量。为了做到这一点,自动编码器(AE)对来自具有固定参数的预训练分类器的梯度进行了微调。我们比较了在一组图像分类器(AlexNet,VGG 16,ResNet-50和Inception~v3)上进行微调的AE的重建样本,并发现了实质性差异。AE根据反向传播梯度中编码的信息,学习保留输入空间的哪些方面,忽略哪些方面。当一个分类器的信号被第二个分类器使用时,测量精度的变化,出现了全阶关系。该顺序直接取决于每个分类器的输入信号,但与分类精度或网络大小无关。通过测量原始图像和来自不同微调AE的自动编码重建之间的归一化互信息,提供了这种现象的进一步证据。这些发现在神经网络理解领域开辟了新天地,为推理、调试和解释其结果开辟了新途径。我们在文献中提出了四个具体的例子,现在可以用模型使用的输入信号来解释观测结果。

英文摘要:

We propose a novel way to measure and understand convolutional neural networks by quantifying the amount of input signal they let in. To do this, an autoencoder (AE) was fine-tuned on gradients from a pre-trained classifier with fixed parameters. We compared the reconstructed samples from AEs that were fine-tuned on a set of image classifiers (AlexNet, VGG16, ResNet-50, and Inception~v3) and found substantial differences. The AE learns which aspects of the input space to preserve and which ones to ignore, based on the information encoded in the backpropagated gradients. Measuring the changes in accuracy when the signal of one classifier is used by a second one, a relation of total order emerges. This order depends directly on each classifier's input signal but it does not correlate with classification accuracy or network size. Further evidence of this phenomenon is provided by measuring the normalized mutual information between original images and auto-encoded reconstructions from different fine-tuned AEs. These findings break new ground in the area of neural network understanding, opening a new way to reason, debug, and interpret their results. We present four concrete examples in the literature where observations can now be explained in terms of the input signal that a model uses.

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。

↑