Search

Fail-operational systems are a prerequisite for autonomous driving. Without a driver who can act as a fallback solution in a critical failure scenario, the system has to be able to mitigate failures on its own and keep critical applications operational. To reduce redundancy cost, graceful degradation can be applied by repurposing hardware resources at run-time. Critical applications can be kept operational by starting passive backups and shutting down non-critical applications instead to make sufficient resources available. In order to design such systems efficiently, the degradation effects on reliability and cost savings have to be analyzed. In this paper we present our approach to formally analyze the impact of graceful degradation on the reliability of critical and non-critical applications. We then quantify the effect of graceful degradation on the reliability of both critical and non-critical applications in distributed automotive systems and compare the achieved cost reduction with conventional redundancy approaches. In our experiments redundancy overhead could be reduced by 80% compared to active redundancy in a scenario with a balanced mix of critical and non-critical applications using our graceful degradation approach Overall, we present a detailed reliability and cost analysis of graceful degradation in distributed automotive systems. Our findings confirm that using graceful degradation can tremendously reduce cost compared to conventional redundancy approaches with no negative impact on the redundancy of critical applications if a reliability reduction of non-critical applications can be accepted. Our results show that a trade-off between the impact of the degradation on the reliability of non-critical applications and cost reduction has to be made.

A variety of methods exist to explain image classification models. However, whether they provide any benefit to users over simply comparing various inputs and the model's respective predictions remains unclear. We conducted a user study (N=240) to test how such a baseline explanation technique performs against concept-based and counterfactual explanations. To this end, we contribute a synthetic dataset generator capable of biasing individual attributes and quantifying their relevance to the model. In a study, we assess if participants can identify the relevant set of attributes compared to the ground-truth. Our results show that the baseline outperformed concept-based explanations. Counterfactual explanations from an invertible neural network performed similarly as the baseline. Still, they allowed users to identify some attributes more accurately. Our results highlight the importance of measuring how well users can reason about biases of a model, rather than solely relying on technical evaluations or proxy tasks. We open-source our study and dataset so it can serve as a blue-print for future studies. For code see, https://github.com/berleon/do_users_benefit_from_interpretable_vision, Comment: Published at ICLR 2022

A growing number of approaches exist to generate explanations for image classification. However, few of these approaches are subjected to human-subject evaluations, partly because it is challenging to design controlled experiments with natural image datasets, as they leave essential factors out of the researcher's control. With our approach, researchers can describe their desired dataset with only a few parameters. Based on these, our library generates synthetic image data of two 3D abstract animals. The resulting data is suitable for algorithmic as well as human-subject evaluations. Our user study results demonstrate that our method can create biases predictive enough for a classifier and subtle enough to be noticeable only to every second participant inspecting the data visually. Our approach significantly lowers the barrier for conducting human subject evaluations, thereby facilitating more rigorous investigations into interpretable machine learning. For our library and datasets see, https://github.com/mschuessler/two4two/, Comment: 6 pages, 3 figures, presented at the ICLR 2021 RAI workshop

A typical strategy of realizing an adiabatic change of a many-particle system is to vary parameters very slowly on a time scale $t_\text{r}$ much larger than intrinsic equilibration time scales. In the ideal case of adiabatic state preparation, $t_\text{r} \to \infty$, the entropy production vanishes. In systems with conservation laws, the approach to the adiabatic limit is hampered by hydrodynamic long-time tails, arising from the algebraically slow relaxation of hydrodynamic fluctuations. We argue that the entropy production $\Delta S$ of a diffusive system at finite temperature in one or two dimensions is governed by hydrodynamic modes resulting in $\Delta S \sim 1/\sqrt{t_\text{r}}$ in $d=1$ and $\Delta S \sim \ln(t_\text{r})/t_\text{r}$ in $d=2$. In higher dimensions, entropy production is instead dominated by other high-energy modes with $\Delta S \sim 1/t_\text{r}$. In order to verify the analytic prediction, we simulate the non-equilibrium dynamics of a classical two-component gas with point-like particles in one spatial dimension and examine the total entropy production as a function of $t_\text{r}$., Comment: 13 pages, 5 figures

Convolutional neural networks (CNNs) offer great machine learning performance over a range of applications, but their operation is hard to interpret, even for experts. Various explanation algorithms have been proposed to address this issue, yet limited research effort has been reported concerning their user evaluation. In this paper, we report on an online between-group user study designed to evaluate the performance of "saliency maps" - a popular explanation algorithm for image classification applications of CNNs. Our results indicate that saliency maps produced by the LRP algorithm helped participants to learn about some specific image features the system is sensitive to. However, the maps seem to provide very limited help for participants to anticipate the network's output for new images. Drawing on our findings, we highlight implications for design and further research on explainable AI. In particular, we argue the HCI and AI communities should look beyond instance-level explanations., Comment: 10 pages, 6 figures, accepted long paper for ACM IUI 2020

The use of complex machine learning models can make systems opaque to users. Machine learning research proposes the use of post-hoc explanations. However, it is unclear if they give users insights into otherwise uninterpretable models. One minimalistic way of explaining image classifications by a deep neural network is to show only the areas that were decisive for the assignment of a label. In a pilot study, 20 participants looked at 14 of such explanations generated either by a human or the LIME algorithm. For explanations of correct decisions, they identified the explained object with significantly higher accuracy (75.64% vs. 18.52%). We argue that this shows that explanations can be very minimalistic while retaining the essence of a decision, but the decision-making contexts that can be conveyed in this manner is limited. Finally, we found that explanations are unique to the explainer and human-generated explanations were assigned 79% higher trust ratings. As a starting point for further studies, this work shares our first insights into quality criteria of post-hoc explanations.

Two- or three-dimensional metals are usually well described by weakly interacting, fermionic quasiparticles. This concept breaks down in one dimension due to strong Coulomb interactions. There, low-energy electronic excitations are expected to be bosonic collective modes, which fractionalize into independent spin and charge density waves. Experimental research on one-dimensional metals is still hampered by their difficult realization, their limited accessibility to measurements, and by competing or obscuring effects such as Peierls distortions or zero bias anomalies. Here we overcome these difficulties by constructing a well-isolated, one-dimensional metal of finite length present in MoS$_2$ mirror twin boundaries. Using scanning tunneling spectroscopy we measure the single-particle density of the interacting electron system as a function of energy and position in the 1D box. Comparison to theoretical modeling provides unambiguous evidence that we are observing spin-charge separation in real space., Comment: 13 pages, 8 figures

We study the temperature-dependent quantum correction to conductivity due to the interplay of spin density fluctuations and weak disorder for a two-dimensional metal near an antiferromagnetic (AFM) quantum critical point. AFM spin density fluctuations carry large momenta around the ordering vector $\mathbf{Q}$ and, at lowest order of the spin-fermion coupling, only scatter electrons between "hot spots" of the Fermi surface which are connected by $\mathbf{Q}$. Earlier, it was seen that the quantum interference between AFM spin density fluctuations and soft diffusive modes of the disordered metal is suppressed, a consequence of the large-momentum scattering. The suppression of this interference results in a non-singular temperature dependence of the corresponding interaction correction to conductivity. However, at higher order of the spin-fermion coupling, electrons on the entire Fermi surface can be scattered successively by two spin density fluctuations and, in total, suffer a small momentum transfer. This higher-order process can be described by composite modes which carry small momenta. We show that the interference between formally subleading composite modes and diffusive modes generates singular interaction corrections which ultimately dominate over the non-singular first-order correction at low temperatures. We derive an effective low-energy theory from the spin-fermion model which includes the above-mentioned higher-order process implicitly and show that for weak spin-fermion coupling the small-momentum transfer is mediated by a composite propagator. Employing the conventional diagrammatic approach to impurity scattering, we find the correction $\delta \sigma \sim +\ln^2 T$ for temperatures above an exponentially small crossover scale., Comment: 13 pages, 7 figures. Published version