Physics-Informed Neural Networks (PINNs) have emerged as a robust framework for solving Partial Differential Equations (PDEs) by approximating their solutions via neural networks and imposing physics-based constraints on the loss function. Traditionally, Multilayer Perceptrons (MLPs) have been the neural network of choice, with significant progress made in optimizing their training. Recently, Kolmogorov-Arnold Networks (KANs) were introduced as a viable alternative, with the potential of offering better interpretability and efficiency while requiring fewer parameters. In this paper, we present a fast JAX-based implementation of grid-dependent Physics-Informed Kolmogorov-Arnold Networks (PIKANs) for solving PDEs, achieving up to 84 times faster training times than the original KAN implementation. We propose an adaptive training scheme for PIKANs, introducing an adaptive state transition technique to avoid loss function peaks between grid extensions, and a methodology for designing PIKANs with alternative basis functions. Through comparative experiments, we demonstrate that the adaptive features significantly enhance solution accuracy, decreasing the L2 error relative to the reference solution by up to 43.02%. For the studied forward PDE problems, our methodology approaches or surpasses the results obtained from architectures that utilize up to 8.5 times more parameters, highlighting the potential of adaptive, grid-dependent PIKANs as a superior alternative in scientific and engineering applications.
We present James Webb Space Telescope MIRI data of the inner ∼3 × 2 kpc2 of the galaxy IC 5063, in which the jets of a supermassive black hole interact with the gaseous disk they are crossing. Jet-driven outflows were known to be initiated along or near the jet path and to modify the stability of molecular clouds, possibly altering their star formation properties. The MIRI data, of unprecedented resolution and sensitivity in the infrared, now reveal that there are more than 10 discrete regions with outflows, nearly doubling the number of such known regions. Outflows exist near the radio lobes, at the nucleus, in a biconical structure perpendicular to the jet, and in a bubble moving against the disk. In some of them, velocities above escape velocity are observed. Stratification is also observed, with higher ionization or excitation gas attaining higher velocities. More outflows and bow shocks, found further away from the nucleus than the radio lobes, in regions without significant radio emission, reveal the existence of past or weak radio jets that interacted with the interstellar medium. The coincidence of the bow shocks with the optical extended emission line region (EELR) suggests that the jets also contributed to the gas ionization. Maps of the H2 gas excitation temperature, Tex, indicate that the molecular gas is most excited in regions with radio emission. There, Tex is >100 K higher than in the EELR interior. We argue that a combination of jet-related shocks and cosmic rays is likely responsible for this excess molecular gas excitation.
Rolling element bearings are critical components of rotating machinery, with their performance directly influencing the efficiency and reliability of industrial systems. At the same time, bearing faults are a leading cause of machinery failures, often resulting in costly downtime, reduced productivity, and, in extreme cases, catastrophic damage. This study presents a methodology that utilizes Kolmogorov–Arnold Networks—a recent deep learning alternative to Multilayer Perceptrons. The proposed method automatically selects the most relevant features from sensor data and searches for optimal hyper-parameters within a single unified approach. By using shallow network architectures and fewer features, the resulting models are lightweight, easily interpretable, and practical for real-time applications. Validated on two widely recognized datasets for bearing fault diagnosis, the framework achieved perfect F1-Scores for fault detection and high performance in fault and severity classification tasks, including 100% F1-Scores in most cases. Notably, it demonstrated adaptability by handling diverse fault types, such as imbalance and misalignment, within the same dataset. The availability of symbolic representations provided model interpretability, while feature attribution offered insights into the optimal feature types or signals for each studied task. These results highlight the framework’s potential for practical applications, such as real-time machinery monitoring, and for scientific research requiring efficient and explainable models.
To evaluate the role of radio activity in galaxy evolution, we designed a large archival CO survey of radio galaxies (RGs) to determine their molecular gas masses at different epochs. We used a sample of 120 RGs representative of the NVSS 1.4 GHz survey, when flux limited at 0.4 Jy. Of those, 66 galaxies belonged to the ALMA Radio-source Catalogue (ARC) of calibrators and had spectral window tunings around CO (1–0), (2–1), (3–2), or (4–3). We reduced their ALMA data, determined their H2 mass contents, and combined the results with similar results for the remaining 54 galaxies from the literature. We found that, while at all epochs the majority of RGs have undetectable reservoirs, there is a rapid increase in the H2 mass content of the CO-detected RGs with z. At 1 < z < 2.5, one-fourth of the RGs have at least as much molecular gas as simulations would indicate for a typical halo mass of that epoch. These galaxies plausibly have “normal” or even starbursty hosts. Overall, reservoirs of 107 ≲ MH2 ≲ 1010M⊙ are seen at z < 0.3, and 1010 ≲ MH2 ≲ 1012M⊙ at z> 1. Taking into account the completeness correction of the sample, we created the corresponding H2 mass functions at 0.005 < z < 0.3 and 1 < z < 2.5. The local mass function reveals that the number density of low-z RGs with detectable molecular gas reservoirs is only a little lower (a factor of ∼4) than that of pure (or little star-forming) type 1 and 2 AGN in simulations. At 1 < z < 2.5, there is a significant decrease in the number density of high-z RGs due to the rarity of bright radio galaxies. An estimate for the missing faint RGs would, nonetheless, bring populations close again. Finally, we find that the volume density of molecular gas locked up in the brightest 1/5000–1/7000 RGs is similar in the examined z bins. This result likely indicates that the inflow rate on one hand and the star-formation depletion rate plus the jet-driven expulsion rate on the other hand counteract each other in the most luminous RGs of each epoch.
We report the detection of outflowing molecular gas at the center of the nearby radio galaxy NGC6328 (z = 0.014), which has a gigahertz-peaked spectrum radio core and a compact (2 pc) young double radio lobe tracing jet. Utilizing Atacama Large Millimeter/submillimeter Array (ALMA) CO(3 − 2) and CO(2 − 1) observations, as well as a novel code developed to fit the 3D gas distribution and kinematics, to study the molecular gas kinematics, we find that the bulk of the gas is situated within a highly warped disk structure, most likely the result of a past merger event. Our analysis further uncovers, within the inner regions of the gas distribution (R < 300 pc) and at a position angle aligning with that of the radio jet (150°), the existence of two anti-diametric molecular gas structures kinematically detached from the main disk. These structures most likely trace a jet-induced cold gas outflow with a total lower limit mass of 2 × 106M⊙ mass, corresponding to an outflow rate of 2 M⊙ yr−1 and a kinetic power of 2.7 × 1040 erg s−1. The energy required to maintain such a molecular outflow is aligned with the mechanical power of the jet.