Deep Networks Visualisation and Interpretation
- [ ] The (Un)reliability of saliency methods \cite{kindermans2017reliability}
- [ ] DeepXplore: Automated Whitebox Testing of Deep Learning Systems \cite{pei2017deepxplore}
- [ ] Feature Visualisation \cite{olah2017feature} (new blog from google with fancy images)
- [ ] Visualizing the loss landscape of neural nets. \cite{li2017visualizing}
-----------------------------------------------------------------------------------------------------------------------------------------------------
Notes
- [Bernhard] An online book about interpretable machine learning - mentioned on 07/12/2017: https://christophm.github.io/interpretable-ml-book/
-----------------------------------------------------------------------------------------------------------------------------------------------------
Previous Papers
2018
- [x] Deep Adversarial Networks for Biomedical Image Segmentation Utilizing Unannotated Images \cite{zhang2017deep}
2017
- [x] CNN Fixations: An unraveling approach to visualize the discriminative image regions \cite{Mopuri2017}
- [x] Visualizing Deep Neural Network Decisions: Prediction Difference Analysis \cite{Zintgraf2017}
- [x] Soft Proposal Networks for Weakly Supervised Object Localization \cite{Zhu2017}
- [x] Self-supervised Learning for Spinal MRIs \cite{Jamaludin_2017}
- [x] Visual Feature Attribution using Wasserstein GANs \cite{1711.08998}
- [x] Non-local Neural Networks \cite{1711.07971} - from facebookAI - Achieves best results in video classification, object segmentation and pose estimation - Check Reddit discussion
- [x] Distilling a Neural Network Into a Soft Decision Tree \cite{1711.09784} - describes a way of using a trained neural net to create a type of soft decision tree that generalizes better than one learned directly from the training data - Reddit
-----------------------------------------------------------------------------------------------------------------------------------------------------