Interpretability in Computer Vision and Natural Language Processing

Interpretability in CV:

• Evaluating the visualization of what a Deep Neural Network has learned, Layer-wise Relevance Propagation techniques, better than sensitivity and deconvolution methods.
• Interpreting CNN Knowledge Via An Explanatory Graph, AAAI 2018. Songchun Zhu’s group.
• Interpretable CNN, CVPR 2018, Songchun Zhu’s group.
• Network Dissection: Quantifying Interpretability of Deep Visual Representations, TPAMI 2018. Bolei Zhou’s work.

Interpretability in NLP, all papers from Gramham Neubig’s group:

• Multi-space Variational Encoder-Decoders, ACL 2017.
• StructVAE: Tree-structured Latent Variable Models for Semi-supervised Semantic Parsing (acl. 18), ACL 2018. code.
• Unsupervised Learning of Syntactic Structure w/ Invertible Neural Projections, EMNLP 2018. code.
• Evaluating neural network explanation methods using hybrid documents and morphosyntactic agreement, ACL 2018.

---