Present-day federated learning (FL) systems deployed over edge networks have to consistently deal with a large number of workers with high degrees of heterogeneity in data and/or computing capabilities. This diverse set of workers necessitates the development of FL algorithms that allow: (1) flexible worker participation that grants the workers' capability to engage in training at will, (2) varying number of local updates (based on computational resources) at each worker along with asynchronous communication with the server, and (3) heterogeneous data across workers. To address these challenges, in this work, we propose a new paradigm in FL called "Anarchic Federated Learning" (AFL). In stark contrast to conventional FL models, each worker in AFL has complete freedom to choose i) when to participate in FL, and ii) the number of local steps to perform in each round based on its current situation (e.g., battery level, communication channels, privacy concerns). However, AFL also introduces significant challenges in algorithmic design because the server needs to handle the chaotic worker behaviors. Toward this end, we propose two Anarchic FedAvg-like algorithms with two-sided learning rates for both cross-device and cross-silo settings, which are named AFedAvg-TSLR-CD and AFedAvg-TSLR-CS, respectively. For general worker information arrival processes, we show that both algorithms retain the highly desirable linear speedup effect in the new AFL paradigm. Moreover, we show that our AFedAvg-TSLR algorithmic framework can be viewed as a meta-algorithm for AFL in the sense that they can utilize advanced FL algorithms as worker-and/or server-side optimizers to achieve enhanced performance under AFL. We validate the proposed algorithms with extensive experiments on real-world datasets.
The lowest 3 tune-out wavelengths of the four alkaline-earth atoms, Be, Mg, Ca and Sr are determined from tabulations of matrix elements produced from large first principles calculations. The tune-out wavelengths are located near the wavelengths for 3 P o 1 and 1 P o 1 excitations. The measurement of the tune-out wavelengths could be used to establish a quantitative relationship between the oscillator strength of the transition leading to existence of the tune-out wavelength and the dynamic polarizability of the atom at the tune-out frequency. The longest tune-out wavelengths for Be, Mg, Ca, Sr, Ba and Yb are 454.9813 nm, 457.2372 nm, 657.446 nm, 689.200 nm, 788.875 nm and 553.00 nm respectively.
A relativistic description of the structure of heavy alkali-metal atoms and alkali-like ions using S-spinors and L-spinors is developed. The core wave function is defined by a Dirac-Fock calculation using an S-spinor basis. The S-spinor basis is then supplemented with a large set of L-spinors for calculation of the valence wave function in a frozen-core model. The numerical stability of the L-spinor approach is demonstrated by computing the energies and decay rates of several low-lying hydrogen eigenstates, along with the polarizabilities of a Z = 60 hydrogenic ion. The approach is then applied to calculate the dynamic polarizabilities of the 5s, 4d, and 5p states of Sr +. The magic wavelengths at which the Stark shifts between different pairs of transitions are 0 are computed. Determination of the magic wavelengths for the 5s → 4d 3
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.