Latest Research

Can we build neural networks whose structure and computational abilities match a real brain? We are not quite there yet, but our new work shows a strategy for getting closer to this goal.

Synchrotron sources produce intense X-rays that are indispensable for many fields of modern science. The essential information, which is often buried in an ocean of experimental data, can be extracted using machine learning. But such models still lack important insights about the structure of the systems being studied. By allowing experimentalists to provide such insights to adaptive neural networks, we were able to instantly obtain more accurate results.

Online learning platforms are popular tools for acquiring new knowledge on our own. However, these platforms have significant shortcomings. We present a new algorithm allowing us to trace the knowledge of learners more accurately, creating opportunities for empowerment by adapting the learning process to their personalized needs.

More than one million papers are published every year in the field of biomedicine and life sciences – an overwhelming volume. To help navigate through the literature, we created a map of the entire landscape of biomedical research using machine learning tools, in the form of an interactive website that allows you to browse 20 million papers simultaneously and see connections between them.

When children develop into adults, how they learn changes a lot. While children show a lot of random behaviour, adults perform more goal-directed actions. An influential theory describes these changes as being similar to the behaviour of an optimisation algorithm commonly used in machine learning. This empirical test shows that there are striking similarities but also important differences between human development and machine learning algorithms.

Deep learning algorithms are very good at recognizing specific objects (e.g. a dog, a car) within an image ​(known as image classifiers)​. But how do they actually do that? Most often the mechanisms underlying an algorithm’s decision remain opaque. What if we could explain any ​such ​black-box algorithm intuitively and, by doing so, even learn from it?​

Understanding the 3D nature of our world is key to many applications in augmented and virtual reality and simulation. But 3D training data is difficult to obtain. Hence, we develop an algorithm to create 3D graphics that can be trained with 2D images alone. By designing our algorithm such that it can represent 3D data efficiently, we keep the computational cost manageable while moving from 2D images to 3D graphics.

We are no longer baffled by all the tasks algorithms can perform. And apparently, they are now even able to ‘explain’ their output. But is that something we really want?

Diffusion processes in nature are highly complex, and scientists strive to understand them in detail. With a new physics-aware neural network, we were able to model and predict such processes much more precisely than previously possible.

Machines may drive you to work one day, but they currently still fail when faced with unusual situations or noisy data. That’s because machines see the world very differently from humans - but this gap is starting to narrow.

Researchers train a neural network to estimate – in just a few seconds – the precise characteristics of merging black holes based on their gravitational-wave emissions. The network determines the masses and spins of the black holes, where in the sky, at what angle, and how far away from Earth the merger took place.

With hundreds of scientists, we have explored the properties of different neuron types in mice, monkeys and humans using novel experimental techniques and machine learning methods for data analysis. The result is a unique overview of the motor cortex in the brain and its evolution.

Single-molecule localization microscopy is a powerful method to image cellular structures with nanometer resolution. We developed DECODE, a deep learning based analysis algorithm that makes this technique faster and more precise.

Computer models are a great tool to analyze neuronal mechanisms in the brain, but tuning these models to match brain activity has long been a daunting task for scientists. We developed a new machine learning tool that automates this process and used it to develop a simulation environment for a retinal implant.

Spatial soil variability makes a farmer's daily business challenging as it leads to varying growth conditions for field crops. Machine learning can help to map soil properties so that farmers can adapt fertilizing and irrigation management in a time- and cost-efficient way.

The Bayesian formalism can add uncertainty to deep neural networks. But Bayesian deep learning has a reputation as cumbersome and expensive. No longer. Recent results show how to achieve calibrated uncertainty in deep networks efficiently, without affecting their predictive performance.