Search

The availability of representative datasets is an essential prerequisite for many successful artificial intelligence and machine learning models. However, in real life applications these models often encounter scenarios that are inadequately represented in the data used for training. There are various reasons for the absence of sufficient data, ranging from time and cost constraints to ethical considerations. As a consequence, the reliable usage of these models, especially in safety-critical applications, is still a tremendous challenge. Leveraging additional, already existing sources of knowledge is key to overcome the limitations of purely data-driven approaches. Knowledge augmented machine learning approaches offer the possibility of compensating for deficiencies, errors, or ambiguities in the data, thus increasing the generalization capability of the applied models. Even more, predictions that conform with knowledge are crucial for making trustworthy and safe decisions even in underrepresented scenarios. This work provides an overview of existing techniques and methods in the literature that combine data-driven models with existing knowledge. The identified approaches are structured according to the categories knowledge integration, extraction and conformity. In particular, we address the application of the presented methods in the field of autonomous driving., Comment: 111 pages, Added section on Run-time Network Verification

We calculate the leading-order QED radiative corrections to the process $e^- p\rightarrow e^- p l^- l^+ $ in the soft-photon approximation, in two different energy regimes which are of relevance to extract nucleon structure information. In the low-energy region, this process is studied to better constrain the hadronic corrections to precision muonic Hydrogen spectroscopy. In the high-energy region, the beam-spin asymmetry for double virtual Compton scattering allows to directly access the Generalized Parton Distributions. We find that the soft-photon radiative corrections have a large impact on the cross sections and are therefore of paramount importance to extract the nucleon structure information from this process. For the forward-backward asymmetry the radiative corrections are found to affect the asymmetry only around or below the 1\% level, whereas the beam-spin asymmetry is not affected at all in the soft-photon approximation, which makes them gold-plated observables to extract nucleon structure information in both the low- and high-energy regimes., Comment: 32 pages, 25 figures

We evaluate the leading-order QED radiative corrections to the timelike Compton scattering (TCS) process $\gamma p \to l^- l^+ p$. We study these corrections in two energy regimes using different models for the TCS amplitude. In the low-energy regime we calculate the contribution due to the proton and its lowest-energy excitation, the $\Delta(1232)$ resonance. In the high-energy near-forward kinematical regime we calculate the TCS amplitude in a handbag approach in terms of Generalized Parton Distributions (GPDs). On the level of cross sections we find the QED radiative corrections to be in the $5 -10\%$ range in the low-energy regime and around $20\%$ in the high-energy regime. We show that in both the di-lepton forward-backward asymmetry as well as in the photon beam helicity asymmetry these corrections nearly cancel out, making them gold-plated observables to extract the real and imaginary parts of the TCS amplitude. We demonstrate in particular the sensitivity of these asymmetries on GPD parameterizations for a recent CLAS12@JLab TCS experiment., Comment: 16 pages, 13 figures

Thermally activated processes are key to understanding the dynamics of physical systems. Thermal diffusion of (quasi-)particles for instance not only yields information on transport and dissipation processes but is also an exponentially sensitive tool to reveal emergent system properties and enable novel applications such as probabilistic computing. Here we probe the thermal dynamics of topologically stabilized magnetic skyrmion quasi-particles. We demonstrate in a specially tailored low pinning multilayer material system pure skyrmion diffusion that dominates the dynamics. Finally, we analyse the applicability to probabilistic computing by constructing a device, which uses the thermally excited skyrmion dynamics to reshuffle a signal. Such a skyrmion reshuffler is the key missing component for probabilistic computing and by evaluating its performance, we demonstrate the functionality of our device with high fidelity thus enabling probabilistic computing.