Search

We conduct direct numerical simulations to investigate the synchronization of Kolmogorov flows in a periodic box, with a focus on the mechanisms underlying the asymptotic evolution of infinitesimal velocity perturbations, also known as conditional leading Lyapunov vectors. This study advances previous work with a spectral analysis of the perturbation, which clarifies the behaviours of the production and dissipation spectra at different coupling wavenumbers. We show that, in simulations with moderate Reynolds numbers, the conditional leading Lyapunov exponent can be smaller than a lower bound proposed previously based on a viscous estimate. A quantitative analysis of the self-similar evolution of the perturbation energy spectrum is presented, extending the existing qualitative discussion. The prerequisites for obtaining self-similar solutions are established, which include an interesting relationship between the integral length scale of the perturbation velocity and the local Lyapunov exponent. By examining the governing equation for the dissipation rate of the velocity perturbation, we reveal the previously neglected roles of the strain rate and vorticity perturbations and uncover their unique geometrical characteristics.

An important parameter characterising the synchronisation of turbulent flows is the threshold coupling wavenumber. This study investigates the relationship between the threshold coupling wavenumber and the leading Lyapunov vector using large eddy simulations and the SABRA model. Various subgrid-scale stress models, Reynolds numbers, and different coupling methods are examined. A new scaling relation is identified for the leading Lyapunov exponents in large eddy simulations, showing that they approximate those of filtered direct numerical simulations. This interpretation provides a physical basis for results related to the Lyapunov exponents of large eddy simulations, including those related to synchronisation. Synchronization experiments show that the peak wavenumber of the energy spectrum of the leading Lyapunov vector coincides with the threshold coupling wavenumber, in large eddy simulations of box turbulence with standard Smagorinsky or dynamic mixed models as well as in the SABRA model, replicating results from direct numerical simulations of box turbulence. Although the dynamic Smagorinsky model exhibit different behaviour, the totality of the results suggests that the relationship is an intrinsic property of a certain class of chaotic systems. We also confirm that conditional Lyapunov exponents characterize the synchronization process in indirectly coupled systems as they do in directly coupled ones, with their values insensitive to the nature of the master flow. These findings advance the understanding of the role of the Lyapunov vector in the synchronization of turbulence.

The synchronisation between rotating turbulent flows in periodic boxes is investigated numerically. The flows are coupled via a master-slave coupling, taking the Fourier modes with wavenumber below a given value $k_m$ as the master modes. It is found that synchronisation happens when $k_m$ exceeds a threshold value $k_c$, and $k_c$ depends strongly on the forcing scheme. In rotating Kolmogorov flows, $k_c\eta$ does not change with rotation in the range of rotation rates considered, $\eta$ being the Kolmogorov length scale. Even though the energy spectrum has a steeper slope, the value of $k_c\eta$ is the same as that found in isotropic turbulence. In flows driven by a forcing term maintaining constant energy injection rate, synchronisation becomes easier when rotation is stronger. $k_c\eta$ decreases with rotation, and it is reduced significantly for strong rotations when the slope of the energy spectrum approaches $-3$. It is shown that the conditional Lyapunov exponent for a given $k_m$ is reduced by rotation in the flows driven by the second type of forcing, but it increases mildly with rotation for the Kolmogorov flows. The local conditional Lyapunov exponents fluctuate more strongly as rotation is increased, although synchronisation occurs as long as the average conditional Lyapunov exponents are negative. We also look for the relationship between $k_c$ and the energy spectra of the Lyapunov vectors. We find that the spectra always seem to peak around $k_c$, and synchronisation fails when the energy spectra of the conditional Lyapunov vectors have a local maximum in the slaved modes.

Prediction sets capture uncertainty by predicting sets of labels rather than individual labels, enabling downstream decisions to conservatively account for all plausible outcomes. Conformal inference algorithms construct prediction sets guaranteed to contain the true label with high probability. These guarantees fail to hold in the face of distribution shift, which is precisely when reliable uncertainty quantification can be most useful. We propose a novel algorithm for constructing prediction sets with PAC guarantees in the label shift setting. This method estimates the predicted probabilities of the classes in a target domain, as well as the confusion matrix, then propagates uncertainty in these estimates through a Gaussian elimination algorithm to compute confidence intervals for importance weights. Finally, it uses these intervals to construct prediction sets. We evaluate our approach on five datasets: the CIFAR-10, ChestX-Ray and Entity-13 image datasets, the tabular CDC Heart dataset, and the AGNews text dataset. Our algorithm satisfies the PAC guarantee while producing smaller, more informative, prediction sets compared to several baselines.

Deep neural networks (DNN) have become a common sensing modality in autonomous systems as they allow for semantically perceiving the ambient environment given input images. Nevertheless, DNN models have proven to be vulnerable to adversarial digital and physical attacks. To mitigate this issue, several detection frameworks have been proposed to detect whether a single input image has been manipulated by adversarial digital noise or not. In our prior work, we proposed a real-time detector, called VisionGuard (VG), for adversarial physical attacks against single input images to DNN models. Building upon that work, we propose VisionGuard* (VG), which couples VG with majority-vote methods, to detect adversarial physical attacks in time-series image data, e.g., videos. This is motivated by autonomous systems applications where images are collected over time using onboard sensors for decision-making purposes. We emphasize that majority-vote mechanisms are quite common in autonomous system applications (among many other applications), as e.g., in autonomous driving stacks for object detection. In this paper, we investigate, both theoretically and experimentally, how this widely used mechanism can be leveraged to enhance the performance of adversarial detectors. We have evaluated VG* on videos of both clean and physically attacked traffic signs generated by a state-of-the-art robust physical attack. We provide extensive comparative experiments against detectors that have been designed originally for out-of-distribution data and digitally attacked images.