26838 research outputs found
Sort by
Analysis and formal specification of OpenJDK's BitSet: Proof files
Full text linkThis artifact [1] (accompanying our iFM 2023 paper [2]) describes the software we developed that contributed towards our analysis of OpenJDK's BitSet class. This class represents a vector of bits that grows as needed. Our analysis exposed numerous bugs. In our paper, we proposed and compared a number of solutions supported by formal specifications. Full mechanical verification of the BitSet class is not yet possible due to limited support for bitwise operations in KeY and bugs in BitSet. Our artifact contains proofs for a subset of the methods and new proof rules to support bitwise operators
A fast all-optical 3D photoacoustic scanner for clinical vascular imaging
Full text linkThe clinical assessment of microvascular pathologies (in diabetes and in inflammatory skin diseases, for example) requires the visualization of superficial vascular anatomy. Photoacoustic tomography (PAT) scanners based on an all-optical Fabry–Perot ultrasound sensor can provide highly detailed 3D microvascular images, but minutes-long acquisition times have precluded their clinical use. Here we show that scan times can be reduced to a few seconds and even hundreds of milliseconds by parallelizing the optical architecture of the sensor readout, by using excitation lasers with high pulse-repetition frequencies and by exploiting compressed sensing. A PAT scanner with such fast acquisition minimizes motion-related artefacts and allows for the volumetric visualization of individual arterioles, venules, venous valves and millimetre-scale arteries and veins to depths approaching 15 mm, as well as for dynamic 3D images of time-varying tissue perfusion and other haemodynamic events. In exploratory case studies, we used the scanner to visualize and quantify microvascular changes associated with peripheral vascular disease, skin inflammation and rheumatoid arthritis. Fast all-optical PAT may prove useful in cardiovascular medicine, oncology, dermatology and rheumatology
Diagnosing Bias in Recommender Systems
No full textIn this project, we experiment with detecting popularity bias in recommender systems using the frameworks Lenskit and Cornac, and testing the algorithms UserKNN and BMF
Erratum to "Keeping your best options open with AI-based treatment planning in prostate and cervix brachytherapy”
Full text linkThe publisher regrets that one of the authors is missing in the PDF and web version of the article. Namely, Anton Bouter, who is affiliated with the Centrum Wiskunde & Informatica as described above in this erratum. In table 1, a space is missing after the character “ Sigmoid' in column `Sparing criteria'. In table 3, the `%' signs in first column (DVI) should also be in subscript, similar to how it is written in the first column of Table 4. The supplementary material should state that needle contribution is limited to up to 40% of the total given dwell time (in applicator and needles). Lastly, the article should be seen as a research article and the authors would like to be cited as “L.R.M. Dickhoff and R.J. Scholman et al”. The publisher would like to apologise for any inconvenience caused
Adaptive factorization using linear-chained hash tables
Full text linkWe introduce factorized aggregations and worst-case optimal joins
in DuckDB with an adaptive mechanism that only uses them when
they enhance query performance. This builds on the adoption of a
new hash table design (“Linear-Chained”) for equi-joins. Our first
insight is that the collision-free chains of this new design enable
efficient factorized and worst-case optimal processing. We further
defer the decision to use factorization and worst-case optimal joins
from optimization to runtime. Our second insight is that we can
obtain accurate statistics, even if the join inputs lack these (e.g.
because they are sub-queries or Parquet files), by leveraging runtime
heuristics and constructing efficient on-the-fly sketches, during the
hash join build. Finally, we show that machine learning models
using these metrics can achieve close to optimal performance with a
high accuracy. Furthermore, we propose heuristic-based approaches
that offer comparable performance to these models, while relying
on cheaper to obtain run-time statistics and being more explainable
Recent results on science and innovation related to electrical processes of thunderstorms
No full textLightning is a highly energetic electric discharge process in our atmosphere, evolving in several complex stages. Lightning is recognized as an essential climate variable, as it affects the concentration of greenhouse gases. It also threatens electrical and electronic devices, in particular, on elevated structures like wind turbines, and it endangers aircraft built with modern composite materials with inherently low electric conductivity. During the past decades, our fundamental understanding of atmospheric electricity has continued to evolve. For example, during the past 30 years, discharge processes were discovered in the atmosphere above thunderstorms, the so-called transient luminous events (TLEs) in the stratosphere and mesosphere, and terrestrial gamma-ray flashes (TGFs), accompanied with beams of photons, electrons and positrons, were observed from low orbiting satellites passing over thunderstorms. Lightning-like discharges also appear in plasma and high-voltage technology. The SAINT network was formed to bring the different research fields together. SAINT was the “Science And INnovation of Thunderstorms” Marie Skłodowska-Curie Innovative Training Network of the European Union Horizon 2020 program. From 2017 to 2021, 15 PhD students observed lightning processes from satellites and ground, developed models and conducted laboratory experiments. The project bridged between geophysical research, plasma technology and relevant industries. The paper presents a summary of the findings of the SAINT network collaboration
Scintillator decorrelation for self-supervised x-ray radiograph denoising
Full text linkX-ray radiographs from industrial, medical, and laboratory x-ray equipment can degrade severely due to fast and/or low-dose acquisition, x-ray scatter, and electronic noise from the detector instrument. As a consequence, noise and artifacts propagate into computed tomography (CT) images. Recently, a new class of self-supervised deep learning methods, with Noise2Self and Noise2Void, demonstrated state-of-the-art denoising results on data sets of pixelwise statistically-independent noisy images. These methods, called blind-spot networks (BSNs), are promising for applications where clean training examples or pairs of noisy examples are unavailable. For x-ray imaging, however, the detection principle of x-ray scintillators leads to a spatially-correlated mix of Poisson and Gaussian noise, rendering BSNs ineffective. In this article, we propose and validate a denoising workflow that reverts the correlations by a direct deconvolution with an estimate of the scintillator point-response function . We show that it can restore the denoising performance of Noise2Self, and demonstrate it for dynamic sparse-view CT reconstruction of single-bubble gas-solids fluidized beds using a data set of unpaired noisy radiographs from cesium-iodine scintillator flat-panel detectors
Energy optimization induces predictive-coding properties in a multi-compartment spiking neural network model
Full text linkPredictive coding is a prominent theoretical framework for understanding hierarchical sensory processing in the brain, yet how it could be implemented in networks of cortical neurons is still unclear. While most existing studies have taken a hand-wiring approach to creating microcircuits that match experimental results, recent work in rate-based artificial neural networks revealed that suitable cortical connectivity might result from self-organisation given some fundamental computational principle, such as energy efficiency. As no corresponding approach has studied this in more plausible networks of spiking neurons, we here investigate whether predictive coding properties in a multi-compartment spiking neural network can emerge from energy optimisation. We find that a model trained with an energy objective in addition to a task-relevant objective is able to reconstruct internal representations given top-down expectation signals alone. Additionally, neurons in the energy-optimised model show differential responses to expected versus unexpected stimuli, qualitatively similar to experimental evidence for predictive coding. These findings indicate that predictive-coding-like behaviour might be an emergent property of energy optimisation, providing a new perspective on how predictive coding could be achieved in the cortex
Detecting wing fractures in chickens using deep learning, photographs and computed tomography scanning
Full text linkAnimal welfare monitoring is a key part of veterinary surveillance in every poultry slaughterhouse. Among the animal welfare indicators routinely inspected, the prevalence of wing fractures and soft tissues injuries (e.g. bruises) is particularly relevant, because it is related to acute pain and suffering in injured birds. According to current practice, assessment corresponds to visual examination by animal welfare officers. However, taking into consideration the speed of the production line and limitations associated with human inspection (e.g. different visual perception, subjectivism and fatigue), new more objective and automated techniques are desirable. Therefore, the aim of this study was to assess the applicability of three deep learning classification models to detect fractures and/or bruises based on computed tomography (CT) scans and photographs of the wings. Namely, 1. Model_CT (two categories: 1.BROKEN and 2.NON_BROKEN) detecting fractures based on CT scans, 2.Model_Photo_Fractures (1.FRACTURES and 2.NO_FRACTURES) detecting fractures based on photographs and 3.Model_Photo_Bruises (1.BRUISES and 2.NO_BRUISES) detecting bruises based on photographs. To train, validate and test these models 306 CT scans and 285 photographs were collected. The 3D ResNet34 and 2D EfficientNetV2_s architectures were used for the CT and Photo_Models, respectively. The models reached an accuracy of 98 % (Model_CT), 96 % (Model_Photo_Fractures) and 82 % (Model_Photo_Bruises). All in all, applying deep learning to the combination of CT scanning and photography can help to objectively recognize wing fractures and bruises. Consequently, it might lead to more accurate and objective animal welfare monitoring and ultimately to raised animal welfare standards