Abstract Multi-task learning (MTL) enables a single neural network to solve multiple tasks simultaneously, offering efficiency and improved generalization potential through shared representations. A central challenge in MTL is balancing task-specific losses during training to avoid performance degradation. While uncertainty-based loss weighting (UW) is a popular and competitive approach, we argue that it suffers from several limitations, including overfitting, rigid homoscedastic assumptions, and a lack of theoretical grounding for various loss functions. Therefore, we propose Soft Optimal Uncertainty Weighting (UW-SO), a novel loss weighting method that builds on UW by deriving analytically optimal weights and applying softmax normalization with adaptable temperature parameter, thereby alleviating several of the shortcomings of UW. Through extensive experiments across diverse datasets and architectures, we show that UW-SO achieves superior and robust performance compared to a variety of existing loss weighting methods. Additionally, we provide insights into the effects of temperature selection and propose measures to reduce computational demand.

In this paper, we consider the problem of event prediction with multi-variate time series data consisting of heterogeneous (continuous and categorical) variables. The complex dependencies between the variables combined with asynchronicity and sparsity of the data makes the event prediction problem particularly challenging. Most state-of-art approaches address this either by designing hand-engineered features or breaking up the problem over homogeneous variates. In this work, we formulate the (rare) event prediction task as a classification problem with a novel asymmetric loss function and propose an end-to-end deep learning algorithm over symbolic representations of time-series. Symbolic representations are fed into an embedding layer and a Long Short Term Memory Neural Network (LSTM) layer which are trained to learn discriminative features. We also propose a simple sequence chopping technique to speed-up the training of LSTM for long temporal sequences. Experiments on real-world industrial datasets demonstrate the effectiveness of the proposed approach.

Learned boundary maps are known to outperform handcrafted ones as a basis for the watershed algorithm. We show, for the first time, how to train watershed computation jointly with boundary map prediction. The estimator for the merging priorities is cast as a neural network that is convolutional (over space) and recurrent (over iterations). The latter allows learning of complex shape priors. The method gives the best known seeded segmentation results on the CREMI segmentation challenge.