Search

The interior of a living cell is an active, fluctuating, and crowded environment. Yet, it maintains a high level of coherent organization, which is readily apparent in the intracellular transport network. Membrane-bound compartments called endosomes play a key role in carrying cargo, in conjunction with myriad components including cargo adaptor proteins, membrane sculptors, motor proteins, and the cytoskeleton. These components coordinate to effectively navigate the crowded cell interior and transport cargo to specific intracellular locations, even though the underlying protein interactions and enzymatic reactions exhibit stochastic behavior. A major challenge is to measure, analyze, and understand how, despite the inherent stochasticity of the constituent processes, the collective outcomes show an emergent spatiotemporal order that is precise and robust. This review focuses on this intriguing dichotomy, providing insights into the known mechanisms of noise suppression and noise utilization in intracellular transport processes, and also identifies opportunities for future inquiry., Comment: Review article

We present a multi-wavelength analysis using SMA, JCMT, NOEMA, JWST, HST, and SST of two dusty strongly star-forming galaxies, 850.1 and 850.2, seen through the massive cluster lens A1489. These SMA-located sources both lie at z=4.26 and have bright dust continuum emission, but 850.2 is a UV-detected Lyman-break galaxy, while 850.1 is undetected at <2um, even with deep JWST/NIRCam observations. We investigate their stellar, ISM, and dynamical properties, including a pixel-level SED analysis to derive sub-kpc-resolution stellar-mass and Av maps. We find that 850.1 is one of the most massive and highly obscured, Av~5, galaxies known at z>4 with M*~10^11.8 Mo (likely forming at z>6), and 850.2 is one of the least massive and least obscured, Av~1, members of the z>4 dusty star-forming population. The diversity of these two dust-mass-selected galaxies illustrates the incompleteness of galaxy surveys at z>3-4 based on imaging at <2um, the longest wavelengths feasible from HST or the ground. The resolved mass map of 850.1 shows a compact stellar mass distribution, Re(mass)~1kpc, but its expected evolution to z~1.5 and then z~0 matches both the properties of massive, quiescent galaxies at z~1.5 and ultra-massive early-type galaxies at z~0. We suggest that 850.1 is the central galaxy of a group in which 850.2 is a satellite that will likely merge in the near future. The stellar morphology of 850.1 shows arms and a linear bar feature which we link to the active dynamical environment it resides within., Comment: Submitted to ApJ, comments welcome!

While a wide variety of astrophysical and cosmological phenomena suggest the presence of Dark Matter, all evidence remains via its gravitational effect on the known matter. As such, it is conceivable that this evidence could be explained by a modification to gravitation and/or concepts of inertia. Various formulations of modified gravity exist, each giving rise to several non-canonical outcomes. This motivates us to propose an experiment searching for departures from (quantum) Newtonian predictions in a bipartite setting with gravitational accelerations $\lesssim 10^{-10}$ m/s$^2$, i.e., where the effective force needs to be stronger than Newtonian to account for the Dark Matter effects. Since quantum particles naturally source weak gravitation, their non-relativistic dynamics offers opportunities to test this small acceleration regime. We show that two nearby mesoscopic quantum masses accumulate significantly larger entanglement in modified gravity models, such as the Modified Newtonian Dynamics. Our calculations include Casimir-Polder forces as well as tidal effects next to the surface of the earth, and confirm that entanglement is observable within the limits imposed by environmental decoherence. We demonstrate how the temperature can be fine-tuned such that modified gravity is certified simply by witnessing the entanglement generated from uncorrelated thermal states, eliminating the need for precise noise characterization. Overall, the required parameters could be realized in a tabletop experiment., Comment: Accepted for publication as a Letter in Physical Review D

We propose an optimal destination scheduling scheme to improve the physical layer security (PLS) of a power-line communication (PLC) based Internet-of-Things system in the presence of an eavesdropper. We consider a pinhole (PH) architecture for a multi-node PLC network to capture the keyhole effect in PLC. The transmitter-to-PH link is shared between the destinations and an eavesdropper which correlates all end-to-end links. The individual channel gains are assumed to follow independent log-normal statistics. Furthermore, the additive impulsive noise at each node is modeled by an independent Bernoulli-Gaussian process. Exact computable expressions for the average secrecy capacity (ASC) and the probability of intercept (POI) performance over many different networks are derived. Approximate closed-form expressions for the asymptotic ASC and POI are also provided. We find that the asymptotic ASC saturates to a constant level as transmit power increases. We observe that the PH has an adverse effect on the ASC. Although the shared link affects the ASC, it has no effect on the POI. We show that by artificially controlling the impulsive to background noise power ratio and its arrival rate at the receivers, the secrecy performance can be improved.

Single crystals of non-centrosymmetric $s$-wave superconductor LaPt$_{0.88}$Si$_{1.12}$ have been grown by the Czochralski (Cz) technique, whose crystal structure is described by the space group $I4{_1}md$ at ambient conditions. The inter-site mixing between platinum and silicon is confirmed by both single-crystal x-ray diffraction (SXRD) and electron probe micro-analyzer (EPMA). The disordered material exhibits a lower superconducting (SC) transition temperature $T_c$ at 2.02 K as opposed to the highest value of 3.9 K reported in polycrystalline LaPtSi without inter-site mixing. From specific heat, the Sommerfeld coefficient ($\gamma$) is estimated to be 7.85 mJ/mol K$^2$, which is much larger than the values reported for the samples exhibiting higher $T_c$. This is unprecedented as $T_c$ seems to decrease with increase in the electron density of states (DOS) at the Fermi energy and thus $\gamma$. The present work reports on the anomalous behaviour of SC and normal state properties of LaPt$_{x}$Si$_{2-x}$, presumably caused due to the existence of non-trivial topological bands.

A centerpiece of the ever-popular reinforcement learning from human feedback (RLHF) approach to fine-tuning autoregressive language models is the explicit training of a reward model to emulate human feedback, distinct from the language model itself. This reward model is then coupled with policy-gradient methods to dramatically improve the alignment between language model outputs and desired responses. In this work, we adopt a novel perspective wherein a pre-trained language model is itself simultaneously a policy, reward function, and transition function. An immediate consequence of this is that reward learning and language model fine-tuning can be performed jointly and directly, without requiring any further downstream policy optimization. While this perspective does indeed break the traditional agent-environment interface, we nevertheless maintain that there can be enormous statistical benefits afforded by bringing to bear traditional algorithmic concepts from reinforcement learning. Our experiments demonstrate one concrete instance of this through efficient exploration based on the representation and resolution of epistemic uncertainty. In order to illustrate these ideas in a transparent manner, we restrict attention to a simple didactic data generating process and leave for future work extension to systems of practical scale.

Federated Learning (FL) is a promising privacy-preserving distributed learning framework where a server aggregates models updated by multiple devices without accessing their private datasets. Hierarchical FL (HFL), as a device-edge-cloud aggregation hierarchy, can enjoy both the cloud server's access to more datasets and the edge servers' efficient communications with devices. However, the learning latency increases with the HFL network scale due to the increasing number of edge servers and devices with limited local computation capability and communication bandwidth. To address this issue, in this paper, we introduce model pruning for HFL in wireless networks to reduce the neural network scale. We present the convergence analysis of an upper on the l2 norm of gradients for HFL with model pruning, analyze the computation and communication latency of the proposed model pruning scheme, and formulate an optimization problem to maximize the convergence rate under a given latency threshold by jointly optimizing the pruning ratio and wireless resource allocation. By decoupling the optimization problem and using Karush Kuhn Tucker (KKT) conditions, closed-form solutions of pruning ratio and wireless resource allocation are derived. Simulation results show that our proposed HFL with model pruning achieves similar learning accuracy compared with the HFL without model pruning and reduces about 50 percent communication cost.