Search

Surrogate model based Evolution Strategies (like the doubly trained surrogate model CMA-ES, DTS-CMA-ES) use a model of the objective function to reduce the number of function evaluations during optimization. This work investigates to use the expected selection weights averaged over the GP posterior distribution as replacement of the fitness and to guide point-selection for evaluation via the variance of the weights. Results obtained on BBOB show that the proposed technique performs on par with current strategies and allows the usage of surrogate models that are invariant to strictly increasing transformations of the function values. However, initial experiments showed that simple modeling of ranks in the GP does lead to worse results than current GP models of the function values.

Feature-based self-explanatory methods explain their classification in terms of human-understandable features. In the medical imaging community, this semantic matching of clinical knowledge adds significantly to the trustworthiness of the AI. However, the cost of additional annotation of features remains a pressing issue. We address this problem by proposing cRedAnno, a data-/annotation-efficient self-explanatory approach for lung nodule diagnosis. cRedAnno considerably reduces the annotation need by introducing self-supervised contrastive learning to alleviate the burden of learning most parameters from annotation, replacing end-to-end training with two-stage training. When training with hundreds of nodule samples and only 1% of their annotations, cRedAnno achieves competitive accuracy in predicting malignancy, meanwhile significantly surpassing most previous works in predicting nodule attributes. Visualisation of the learned space further indicates that the correlation between the clustering of malignancy and nodule attributes coincides with clinical knowledge. Our complete code is open-source available: https://github.com/diku-dk/credanno.

Probabilistic image segmentation encodes varying prediction confidence and inherent ambiguity in the segmentation problem. While different probabilistic segmentation models are designed to capture different aspects of segmentation uncertainty and ambiguity, these modelling differences are rarely discussed in the context of applications of uncertainty. We consider two common use cases of segmentation uncertainty, namely assessment of segmentation quality and active learning. We consider four established strategies for probabilistic segmentation, discuss their modelling capabilities, and investigate their performance in these two tasks. We find that for all models and both tasks, returned uncertainty correlates positively with segmentation error, but does not prove to be useful for active learning.

We present a novel black box optimization algorithm called Hessian Estimation Evolution Strategy. The algorithm updates the covariance matrix of its sampling distribution by directly estimating the curvature of the objective function. This algorithm design is targeted at twice continuously differentiable problems. For this, we extend the cumulative step-size adaptation algorithm of the CMA-ES to mirrored sampling. We demonstrate that our approach to covariance matrix adaptation is efficient by evaluating it on the BBOB/COCO testbed. We also show that the algorithm is surprisingly robust when its core assumption of a twice continuously differentiable objective function is violated. The approach yields a new evolution strategy with competitive performance, and at the same time it also offers an interesting alternative to the usual covariance matrix update mechanism.

Wave energy is a widely available but still largely unexploited energy source, which has not yet reached full commercial development. A common design for a wave energy converter is called a point absorber (or buoy), which either floats on the surface or just below the surface of the water. Since a single buoy can only capture a limited amount of energy, large-scale wave energy production requires the deployment of buoys in large numbers called arrays. However, the efficiency of these arrays is affected by highly complex constructive and destructive intra-buoy interactions. We tackle the multi-objective variant of the buoy placement problem: we are taking into account the highly complex interactions of the buoys, while optimising critical design aspects: the energy yield, the necessary area, and the cable length needed to connect all buoys – while considering realistic wave conditions for the first time, i.e., a real wave spectrum and waves from multiple directions. To make the problem computationally feasible, we use sparse incomplete LU decomposition for solving systems of equations, and caching of integral computations. For the optimisation, we employ modern multi-objective solvers that are customised to the buoy placement problems. We analyse the wave field of final solutions to confirm the quality of the achieved layouts.

We compare a set of convolutional neural network (CNN) architectures for the task of segmenting and detecting human sperm cells in an image taken from a semen sample. In contrast to previous work, samples are not stained or washed to allow for full sperm quality analysis, making analysis harder due to clutter. Our results indicate that training on full images is superior to training on patches when class-skew is properly handled. Full image training including up-sampling during training proves to be beneficial in deep CNNs for pixel wise accuracy and detection performance. Predicted sperm cells are found by using connected components on the CNN predictions. We investigate optimization of a threshold parameter on the size of detected components. Our best network achieves 93.87% precision and 91.89% recall on our test dataset after thresholding outperforming a classical image analysis approach.