This work investigates the challenging problem of domain generalization, i.e., training a model on multi-domain source data such that it can directly generalize to target domains with unknown statistics, and adopts a model-agnostic learning paradigm with gradient-based meta-train and meta-test procedures to expose the optimization to domain shift.
The authors show how causal reasoning can shed new light on scarcity of high-quality annotated data and mismatch between the development dataset and the target environment, and show step-by-step recommendations for future studies.
The experimental results indicate that the proposed framework can successfully train deep SCMs that are capable of all three levels of Pearl's ladder of causation: association, intervention, and counterfactuals, giving rise to a powerful new approach for answering causal questions in imaging applications and beyond.
Stochastic segmentation networks (SSNs) are introduced, an efficient probabilistic method for modelling aleatoric uncertainty with any image segmentation network architecture and outperform state-of-the-art for modelling correlated uncertainty in ambiguous images while being much simpler, more flexible, and more efficient.
A novel cost function for semi-supervised learning of neural networks that encourages compact clustering of the latent space to facilitate separation and can be easily applied to existing networks to enable an effective use of unlabeled data.
The analysis of scanner effects when using machine learning on multi-site neuroimaging data suggests that current approaches to harmonize data are unable to remove scanner-specific bias leading to overly optimistic performance estimates and poor generalization.
Morpho-MNIST is introduced, a framework that aims to answer: "to what extent has my model learned to represent specific factors of variation in the data" and a set of quantifiable perturbations to assess the performance of unsupervised and supervised methods on challenging tasks such as outlier detection and domain adaptation.
This work proposes two deep architectures, an end-to-end synthesis network and a latent feature interpolation network, to predict cardiac segmentation maps from extremely undersampled dynamic MRI data, bypassing the usual image reconstruction stage altogether.
It is shown that cleaning noisy labels mitigates their negative impact on model training, evaluation, and selection and enables correcting labels up to 4 × more effectively than typical random selection in realistic conditions, making better use of experts’ valuable time for improving dataset quality.
It is shown that the proposed Bayesian methods achieve competitive performance when the test images are relatively far from the training data distribution and outperforms when the baseline method is over-parametrised, demonstrating the potential utility of the Bayesian DL methods for assessing the reliability of the reconstructed images.