This article studies the Gram random matrix model G = 1 T Σ T Σ, Σ = σ(W X), classically found in the analysis of random feature maps and random neural networks, where X = [x1, . . . , xT ] ∈ R p×T is a (data) matrix of bounded norm, W ∈ R n×p is a matrix of independent zero-mean unit variance entries, and σ : R → R is a Lipschitz continuous (activation) function -σ(W X) being understood entrywise. By means of a key concentration of measure lemma arising from non-asymptotic random matrix arguments, we prove that, as n, p, T grow large at the same rate, the resolvent Q = (G + γIT ) −1 , for γ > 0, has a similar behavior as that met in sample covariance matrix models, involving notably the moment Φ = T n E[G], which provides in passing a deterministic equivalent for the empirical spectral measure of G. Application-wise, this result enables the estimation of the asymptotic performance of single-layer random neural networks. This in turn provides practical insights into the underlying mechanisms into play in random neural networks, entailing several unexpected consequences, as well as a fast practical means to tune the network hyperparameters.

In this paper, we provide a novel construction of the linear-sized spectral sparsifiers of Batson, Spielman and Srivastava [BSS14]. While previous constructions required Ω(n 4 ) running time [BSS14, Zou12], our sparsification routine can be implemented in almost-quadratic running time O(n 2+ε ). The fundamental conceptual novelty of our work is the leveraging of a strong connection between sparsification and a regret minimization problem over density matrices. This connection was known to provide an interpretation of the randomized sparsifiers of Spielman and Srivastava [SS11] via the application of matrix multiplicative weight updates (MWU) [CHS11,Vis14]. In this paper, we explain how matrix MWU naturally arises as an instance of the Follow-theRegularized-Leader framework and generalize this approach to yield a larger class of updates. This new class allows us to accelerate the construction of linear-sized spectral sparsifiers, and give novel insights on the motivation behind Batson, Spielman and Srivastava [BSS14].

No abstract

To develop a more efficient and optimal artificial kidney, many experimental approaches have been used to study mass transfer inside, outside, and cross hollow fiber membranes with different kinds of membranes, solutes, and flow rates as parameters. However, these experimental approaches are expensive and time consuming. Numerical calculation and computer simulation is an effective way to study mass transfer in the artificial kidney, which can save substantial time and reduce experimental cost. This paper presents a new model to simulate mass transfer in artificial kidney by coupling together shell-side, lumen-side, and transmembrane flows. Darcy's equations were employed to simulate shell-side flow, Navier-Stokes equations were employed to simulate lumen-side flow, and Kedem-Katchalsky equations were used to compute transmembrane flow. Numerical results agreed well with experimental results within 10% error. Numerical results showed the nonuniform distribution of flow and solute concentration in shell-side flow due to the entry/exit effect and Darcy permeability. In the shell side, the axial velocity in the periphery is higher than that in the center. This numerical model presented a clear insight view of mass transfer in an artificial kidney and may be used to help design an optimal artificial kidney and its operation conditions to improve hemodialysis.

Articles you may be interested inStructural and dielectric properties of laser ablated BaTiO3 films deposited over electrophoretically dispersed CoFe2O4 grains Raman spectroscopic study of Na1/2Bi1/2TiO3-x%BaTiO3 single crystals as a function of temperature and compositionThe low Curie temperature (T c ¼ 130 C) of bulk BaTiO 3 greatly limits its applications. In this work, the phase structures of BaTiO 3 nanoparticles with sizes ranging from 2.5 nm to 10 nm were studied at various temperatures by using aberration-corrected transmission electron microscopy (TEM) equipped with an in-situ heating holder. The results implied that each BaTiO 3 nanoparticle was composed of different phases, and the ferroelectric ones were observed in the shells due to the complicated surface structure. The ferroelectric phases in BaTiO 3 nanoparticles remained at 600 C, suggesting a significant increase of T c . Based on the in-situ TEM results and the data reported by others, temperature-size phase diagrams for BaTiO 3 particles and ceramics were proposed, showing that the phase transition became diffused and the T c obviously increased with decreasing size. The present work sheds light on the design and fabrication of advanced devices for high temperature applications.