arXiv subjects

Tobias Huber

Publications and source records attributed to Tobias Huber.

At least 37 records · Page 2Linked to original sources

arXiv2022

Integrating Policy Summaries with Reward Decomposition for Explaining Reinforcement Learning Agents

Explaining the behavior of reinforcement learning agents operating in sequential decision-making settings is challenging, as their behavior is affected by a dynamic environment and delayed rewards. Methods that help users understand the behavior of such agents can roughly be divided into local explanations that analyze specific decisions of the agents and global explanations that convey the general strategy of the agents. In this work, we study a novel combination of local and global explanations for reinforcement learning agents. Specifically, we combine reward decomposition, a local explanation method that exposes which components of the reward function influenced a specific decision, and HIGHLIGHTS, a global explanation method that shows a summary of the agent's behavior in decisive states. We conducted two user studies to evaluate the integration of these explanation methods and their respective benefits. Our results show significant benefits for both methods. In general, we found that the local reward decomposition was more useful for identifying the agents' priorities. However, when there was only a minor difference between the agents' preferences, then the global information provided by HIGHLIGHTS additionally improved participants' understanding.

cs.LG↗

arXiv2022

IBP reduction via Gröbner bases in a rational double-shift algebra

We report on an approach to integration-by-parts reduction based on Gröbner bases. We establish the underlying noncommutative rational double-shift algebra wherein the integration-by-parts relations form a left ideal. We describe in detail the one-loop massless box as an example where we achieved the full reduction to master integrals by means of the Gröbner basis approach, and report on the performance of the implementation. We also identify potential bottlenecks in more complicated examples and elaborate on interesting further directions.

hep-ph↗

arXiv2022

Alterfactual Explanations -- The Relevance of Irrelevance for Explaining AI Systems

Explanation mechanisms from the field of Counterfactual Thinking are a widely-used paradigm for Explainable Artificial Intelligence (XAI), as they follow a natural way of reasoning that humans are familiar with. However, all common approaches from this field are based on communicating information about features or characteristics that are especially important for an AI's decision. We argue that in order to fully understand a decision, not only knowledge about relevant features is needed, but that the awareness of irrelevant information also highly contributes to the creation of a user's mental model of an AI system. Therefore, we introduce a new way of explaining AI systems. Our approach, which we call Alterfactual Explanations, is based on showing an alternative reality where irrelevant features of an AI's input are altered. By doing so, the user directly sees which characteristics of the input data can change arbitrarily without influencing the AI's decision. We evaluate our approach in an extensive user study, revealing that it is able to significantly contribute to the participants' understanding of an AI. We show that alterfactual explanations are suited to convey an understanding of different aspects of the AI's reasoning than established counterfactual explanation methods.

cs.AI↗

arXiv2022

$B\to D \ell ν_\ell$ form factors beyond leading power and extraction of $|V_{cb}|$ and $R(D)$

We investigate the subleading-power corrections to the exclusive $B\to D \ell ν_\ell$ form factors at ${\cal O} (α_s^0)$ in the light-cone sum rules (LCSR) framework by including the two- and three-particle higher-twist contributions from the $B$-meson light-cone distribution amplitudes up to the twist-six accuracy, by taking into account the subleading terms in expanding the hard-collinear charm-quark propagator, and by evaluating the hadronic matrix element of the subleading effective current $\bar q \, γ_μ \, i \not\!\!{D}_\perp / (2 \, m_b) \, h_v$. Employing further the available leading-power results for the semileptonic $B \to D$ form factors at the next-to-leading-logarithmic level and combining our improved LCSR predictions with the recent lattice determinations, we then carry out a comprehensive phenomenological analysis on the semi-leptonic $B\to D \ell ν_\ell$ decay. We extract $|V_{cb}| = \big( 40.2^{+0.6}_{-0.5} {\big |_{\rm th}}\,\, {}^{+1.4}_{-1.4} {\big |_{\rm exp}} \big)\times 10^{-3}$ ($|V_{cb}| = \big( 40.9^{+0.6}_{-0.5} {\big |_{\rm th}}\,\, {}^{+1.0}_{-1.0} {\big |_{\rm exp}} \big)\times 10^{-3}$) using the BaBar (Belle) experimental data, and particularly obtain for the gold-plated ratio $R(D) = 0.302\pm 0.003$.

hep-ph↗

arXiv2022

Some recent developments in nonleptonic $B$ decays

I report on three recent topics from the field of two-body nonleptonic decays of $B_{(s)}$ mesons. The computation of two-loop NNLO corrections to the leading penguin amplitudes in QCD factorisation, puzzles that have emerged in colour-allowed tree-level decays to heavy-light final states, and a combination of QCD factorisation with $SU(3)$ flavour symmetry to estimate the size of weak-annihilation amplitudes.

hep-ph↗

arXiv2022

Estimating QCD-factorization amplitudes through $SU(3)$ symmetry in $B\rightarrow P P$ decays

We estimate the potential size of the weak annihilation amplitudes in QCD factorization as allowed by experimental data. To achieve this goal, we establish a connection between the amplitudes in the QCD factorization and the so-called topological and $SU(3)$-invariant descriptions. Our approach is based purely on the analysis of the tensor structure of the decay amplitudes. By focusing on the processes $B\rightarrow P P$ and considering data from CP asymmetries and branching fractions, we perform a global fit to the $SU(3)$-irreducible quantities assuming minimal theoretical bias. Subsequently we translate the outcome to the QCD factorization decomposition, and find that the most constrained weak annihilation amplitudes are below $4\%$. However, in view of the large uncertainties in several of the experimental input parameters, values up to $30\%$ are allowed in certain cases.

hep-ph↗

arXiv2022

The $Hb\bar{b}$ vertex at four loops and hard matching coefficients in SCET for various currents

We compute the four-loop corrections to the Higgs-bottom vertex within massless QCD and present analytic results for all color structures. The infrared poles of the renormalized form factor agree with the predicted four-loop pattern. Furthermore, we use the results for the Higgs-bottom, photon-quark, and Higgs-gluon form factors to provide hard matching coefficients in soft-collinear effective theory up to four-loop accuracy.

hep-ph↗

arXiv2022

Do Deep Neural Networks Forget Facial Action Units? -- Exploring the Effects of Transfer Learning in Health Related Facial Expression Recognition

In this paper, we present a process to investigate the effects of transfer learning for automatic facial expression recognition from emotions to pain. To this end, we first train a VGG16 convolutional neural network to automatically discern between eight categorical emotions. We then fine-tune successively larger parts of this network to learn suitable representations for the task of automatic pain recognition. Subsequently, we apply those fine-tuned representations again to the original task of emotion recognition to further investigate the differences in performance between the models. In the second step, we use Layer-wise Relevance Propagation to analyze predictions of the model that have been predicted correctly previously but are now wrongly classified. Based on this analysis, we rely on the visual inspection of a human observer to generate hypotheses about what has been forgotten by the model. Finally, we test those hypotheses quantitatively utilizing concept embedding analysis methods. Our results show that the network, which was fully fine-tuned for pain recognition, indeed payed less attention to two action units that are relevant for expression recognition but not for pain recognition.

cs.CV↗

arXiv2022

Numerical optimization of single-mode fiber-coupled single-photon sources based on semiconductor quantum dots

We perform extended numerical studies to maximize the overall photon coupling efficiency of fiber-coupled quantum dot single-photon sources emitting in the near-infrared and telecom regime. Using the finite element method, we optimize the photon extraction and fiber-coupling efficiency of quantum dot single-photon sources based on micromesas, microlenses, circular Bragg grating cavities and micropillars. The numerical simulations which consider the entire system consisting of the quantum dot source itself, the coupling lens, and the single-mode fiber yield overall photon coupling efficiencies of up to 83%. Our work provides objectified comparability of different fiber-coupled single-photon sources and proposes optimized geometries for the realization of practical and highly efficient quantum dot single-photon sources.

quant-ph↗

arXiv2022

Next-to-Next-to-Leading-Order QCD Prediction for the Photon-Pion Form Factor

We accomplish the complete two-loop computation of the leading-twist contribution to the photon-pion transition form factor $γ\, γ^{\ast} \to π^0$ by applying the hard-collinear factorization theorem together with modern multi-loop techniques. The resulting predictions for the form factor indicate that the two-loop perturbative correction is numerically important. We also demonstrate that our results will play a key role in disentangling various models of the twist-two pion distribution amplitude thanks to the envisaged precision at Belle II.

hep-ph↗

arXiv2022

Benchmarking Perturbation-based Saliency Maps for Explaining Atari Agents

One of the most prominent methods for explaining the behavior of Deep Reinforcement Learning (DRL) agents is the generation of saliency maps that show how much each pixel attributed to the agents' decision. However, there is no work that computationally evaluates and compares the fidelity of different saliency map approaches specifically for DRL agents. It is particularly challenging to computationally evaluate saliency maps for DRL agents since their decisions are part of an overarching policy. For instance, the output neurons of value-based DRL algorithms encode both the value of the current state as well as the value of doing each action in this state. This ambiguity should be considered when evaluating saliency maps for such agents. In this paper, we compare five popular perturbation-based approaches to create saliency maps for DRL agents trained on four different Atari 2600 games. The approaches are compared using two computational metrics: dependence on the learned parameters of the agent (sanity checks) and fidelity to the agent's reasoning (input degradation). During the sanity checks, we encounter issues with one approach and propose a solution to fix these issues. For fidelity, we identify two main factors that influence which saliency approach should be chosen in which situation.

cs.LG↗

arXiv2021

The pion-photon transition form factor at two loops in QCD

We report on the fully analytic calculation of the leading-power contribution to the photon-pion transition form factor $γ\, γ^{\ast} \to π^0$ at two loops in QCD. The applied techniques are based on hard-collinear factorization, together with modern multi-loop methods. We focus both, on the technical details such as the treatment of evanescent operators, and the phenomenological implications. Our results indicate that the two-loop correction is numerically comparable to the one-loop effect in the same kinematic domain. We also demonstrate that our results will play a key role in disentangling various models for the twist-two pion distribution amplitude thanks to the envisaged precision at Belle II.

hep-ph↗

arXiv2021

Chiral transport of hot carriers in graphene in the quantum Hall regime

Photocurrent (PC) measurements can reveal the relaxation dynamics of photo-excited hot carriers beyond the linear response of conventional transport experiments, a regime important for carrier multiplication. In graphene subject to a magnetic field, PC measurements are able to probe the existence of Landau levels with different edge chiralities which is exclusive to relativistic electron systems. Here, we report the accurate measurement of PC in graphene in the quantum Hall regime. Prominent PC oscillations as a function of gate voltage on samples' edges are observed. These oscillation amplitudes form an envelope which depends on the strength of the magnetic field, as does the PCs' power dependence and their saturation behavior. We explain these experimental observations through a model using optical Bloch equations, incorporating relaxations through acoustic-, optical- phonons and Coulomb interactions. The simulated PC agrees with our experimental results, leading to a unified understanding of the chiral PC in graphene at various magnetic field strengths, and providing hints for the occurrence of a sizable carrier multiplication.

cond-mat.mes-hall↗

arXiv2021

Dynamic Difficulty Adjustment in Virtual Reality Exergames through Experience-driven Procedural Content Generation

Virtual Reality (VR) games that feature physical activities have been shown to increase players' motivation to do physical exercise. However, for such exercises to have a positive healthcare effect, they have to be repeated several times a week. To maintain player motivation over longer periods of time, games often employ Dynamic Difficulty Adjustment (DDA) to adapt the game's challenge according to the player's capabilities. For exercise games, this is mostly done by tuning specific in-game parameters like the speed of objects. In this work, we propose to use experience-driven Procedural Content Generation for DDA in VR exercise games by procedurally generating levels that match the player's current capabilities. Not only finetuning specific parameters but creating completely new levels has the potential to decrease repetition over longer time periods and allows for the simultaneous adaptation of the cognitive and physical challenge of the exergame. As a proof-of-concept, we implement an initial prototype in which the player must traverse a maze that includes several exercise rooms, whereby the generation of the maze is realized by a neural network. Passing those exercise rooms requires the player to perform physical activities. To match the player's capabilities, we use Deep Reinforcement Learning to adjust the structure of the maze and to decide which exercise rooms to include in the maze. We evaluate our prototype in an exploratory user study utilizing both biodata and subjective questionnaires.

cs.HC↗

arXiv2021

GANterfactual - Counterfactual Explanations for Medical Non-Experts using Generative Adversarial Learning

With the ongoing rise of machine learning, the need for methods for explaining decisions made by artificial intelligence systems is becoming a more and more important topic. Especially for image classification tasks, many state-of-the-art tools to explain such classifiers rely on visual highlighting of important areas of the input data. Contrary, counterfactual explanation systems try to enable a counterfactual reasoning by modifying the input image in a way such that the classifier would have made a different prediction. By doing so, the users of counterfactual explanation systems are equipped with a completely different kind of explanatory information. However, methods for generating realistic counterfactual explanations for image classifiers are still rare. Especially in medical contexts, where relevant information often consists of textural and structural information, high-quality counterfactual images have the potential to give meaningful insights into decision processes. In this work, we present GANterfactual, an approach to generate such counterfactual image explanations based on adversarial image-to-image translation techniques. Additionally, we conduct a user study to evaluate our approach in an exemplary medical use case. Our results show that, in the chosen medical use-case, counterfactual explanations lead to significantly better results regarding mental models, explanation satisfaction, trust, emotions, and self-efficacy than two state-of-the-art systems that work with saliency maps, namely LIME and LRP.

cs.LG↗

arXiv2021

Optical Charge Injection and Full Coherent Control of Spin-Qubit in the Telecom C-band Emitting Quantum Dot

Solid-state quantum emitters with manipulable spin-qubits are promising platforms for quantum communication applications. Although such light-matter interfaces could be realized in many systems only a few allow for light emission in the telecom bands necessary for long-distance quantum networks. Here, we propose and implement a new optically active solid-state spin-qubit based on a hole confined in a single InAs/GaAs quantum dot grown on an InGaAs metamorphic buffer layer emitting photons in the C-band. We lift the hole spin-degeneracy using an external magnetic field and demonstrate hole injection, initialization, read-out and complete coherent control using picosecond optical pulses. These results showcase a new solid-state spin-qubit platform compatible with preexisting optical fibre networks.

cond-mat.mes-hall↗

arXiv2020

Phenomenology of inclusive $\bar{B}\to X_s \ell^+ \ell^-$ for the Belle II era

With the first data being recorded at Belle II, we are at the brink of a new era in quark flavour physics. The many exciting new opportunities for Belle~II include a full angular analysis of inclusive ${\bar B \to X_{s} \, \ell^+\ell^-}$ which has the potential to reveal new physics, in particular by its interplay with the exclusive $b \to s \ell^+\ell^-$ counterparts studied extensively at LHCb. In this paper, we present fully updated Standard Model predictions for all angular observables necessary for this endeavour. These predictions are tailored to Belle II and include an elaborate study of the treatment of collinear photons which become crucial when aiming for the highest precision. In addition, we present a phenomenological study of the potential for Belle II to reveal possible new physics in the inclusive decay channel, both in an independent manner and in combination with exclusive modes.

hep-ph↗

arXiv2020

A puzzle in $\bar{B}_{(s)}^0 \to D_{(s)}^{(*)+} \lbrace π^-, K^-\rbrace$ decays and extraction of the $f_s/f_d$ fragmentation fraction

We provide updated predictions for the hadronic decays $\bar{B}_s^0\to D_s^{(*)+} π^-$ and $\bar{B}^0\to D^{(*)+} K^-$. They are based on $\mathcal{O}(α_s^2)$ results for the QCD factorization amplitudes at leading power and on recent results for the $\bar{B}_{(s)} \to D_{(s)}^{(*)}$ form factors up to order ${\cal O}(Λ_{\rm QCD}^2/m_c^2)$ in the heavy-quark expansion. We give quantitative estimates of the matrix elements entering the hadronic decay amplitudes at order ${\cal O}(Λ_{\rm QCD}/m_b)$ for the first time. Our results are very precise, and uncover a substantial discrepancy between the theory predictions and the experimental measurements. We explore two possibilities for this discrepancy: non-factorizable contributions larger than predicted by the QCD factorization power counting, and contributions beyond the Standard Model. We determine the $f_s/f_d$ fragmentation fraction for the CDF, D0 and LHCb experiments for both scenarios.

hep-ph↗