Soroosh Baselizadeh

Is critical input information encoded in specific sparse pathways within the neural network? In this work, we discuss the problem of identifying these critical pathways and subsequently leverage them for interpreting the network’s response to an input. We demonstrate that sparse pathways derived from pruning do not necessarily encode critical input information. …The pruning objective – selecting the smallest group of neurons for which the response remains equivalent to the original network – has been previously proposed for identifying critical pathways. As stated, using pruning objective is flawed for obtaining pathways encoding critical input information. To ensure sparse pathways include critical fragments of the encoded input information, we propose pathway selection via neurons’ contribution to the response. We proceed to explain how critical pathways can reveal critical input features. We prove that pathways selected via neuron contribution are locally linear (in an L2-ball), a property that we use for proposing a feature attribution method: “pathway gradient”. We validate our interpretation method using mainstream evaluation experiments. The validation of pathway gradient interpretation method further confirms that selected pathways using neuron contributions correspond to critical input features. More

Ashkan Khakzar, Soroosh Baselizadeh, Saurabh Khanduja, Christian Rupprecht, Seong Tae Kim, and Nassir Navab

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2021 [arXiv][CVF OpenAccess]

Here, we propose to use the “distillation” of features at various layers of an expert network, which is pre-trained on ImageNet, into a simpler cloner network to tackle issues in unsupervised representation learning for anomaly detection/localization. We detect and localize anomalies using the discrepancy between the expert and cloner networks’ intermediate activation values given an input sample. … Unsupervised representation learning has proved to be a critical component of anomaly detection/localization in images. The challenges to learn such a representation are two-fold. Firstly, the sample size is not often large enough to learn a rich generalizable representation through conventional techniques. Secondly, while only normal samples are available at training, the learned features should be discriminative of normal and anomalous samples. Here, we propose to use the “distillation” of features at various layers of an expert network, which is pre-trained on ImageNet, into a simpler cloner network to tackle both issues. We detect and localize anomalies using the discrepancy between the expert and cloner networks’ intermediate activation values given an input sample. We show that considering multiple intermediate hints in distillation leads to better exploitation of the expert’s knowledge and a more distinctive discrepancy between the two networks, compared to utilizing only the last layer activation values. Notably, previous methods either fail in precise anomaly localization or need expensive region-based training. In contrast, with no need for any special or intensive training procedure, we incorporate interpretability algorithms in our novel framework to localize anomalous regions. Despite the striking difference between some test datasets and ImageNet, we achieve competitive or significantly superior results compared to SOTA on MNIST, F-MNIST, CIFAR-10, MVTecAD, Retinal-OCT, and two other medical datasets on both anomaly detection and localization. More

MohammadReza Salehi, Niousha Sadjadi*, Soroosh Baselizadeh*, M. H. Rohban, and Hamid R. Rabiee

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2021 [arXiv][CVF OpenAccess]

The black-box nature of CNNs has sparked many recent works to explain the prediction via input feature attribution methods. However, input these methods merely identify the importance of input regions for the prediction and lack semantic interpretation of model behavior. Here, we first identify the semantics associated with internal units (feature maps) of the network. We proceed to investigate the following questions; …Does a regression model that is only trained with COVID-19 severity scores implicitly learn visual patterns associated with thoracic pathologies? Does a network that is trained on weakly labeled data (e.g. healthy, unhealthy) implicitly learn pathologies? Moreover, we investigate the effect of pretraining and data imbalance on the interpretability of learned features. In addition to the analysis, we propose semantic attribution to semantically explain each prediction. We present our findings using publicly available chest pathologies (CheXpert, NIH ChestX-ray8) and COVID-19 datasets (BrixIA, and COVID-19 chest X-ray segmentation dataset). More

A. Khakzar, S. Musatian, J. Buchberger, I. Valeriano Quiroz, N. Pinger, S. Baselizadeh, S. T. Kim, Nassir Navab

International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2021 [arXiv] [Springer (DOI)] [SharedIt]

A popular family of saliency methods utilize gradient information. In this work, we empirically show that two approaches for handling the gradient information, namely positive aggregation, and positive propagation, (e.g. in methods like GradCAM++ and FullGrad) break these methods. Though these methods reflect visually salient information in the input, they do not explain the model prediction anymore …as the generated saliency maps are insensitive to the predicted output and are insensitive to model parameter randomization. Specifically for methods that aggregate the gradients of a chosen layer such as GradCAM++ and FullGrad, exclusively aggregating positive gradients is detrimental. We further support this by proposing several variants of aggregation methods with positive handling of gradient information. For methods that backpropagate gradient information such as LRP, RectGrad, and Guided Backpropagation, we show the destructive effect of exclusively propagating positive gradient information. More

Ashkan Khakzar, Soroosh Baselizadeh, Nassir Navab

International Conference on Machine Learning (ICML) 2020 - Workshop on Human Interpretability in ML (WHI) [arXiv][Full Proceedings of the Venue]