arXiv subjects

Huan Zhao

Publications and source records attributed to Huan Zhao.

139 records · Page 8Linked to original sources

arXiv2019

Machine Learning Approach for Device-Circuit Co-Optimization of Stochastic-Memristive-Device-Based Boltzmann Machine

A Boltzmann machine whose effective "temperature" can be dynamically "cooled" provides a stochastic neural network realization of simulated annealing, which is an important metaheuristic for solving combinatorial or global optimization problems with broad applications in machine intelligence and operations research. However, the hardware realization of the Boltzmann stochastic element with "cooling" capability has never been achieved within an individual semiconductor device. Here we demonstrate a new memristive device concept based on two-dimensional material heterostructures that enables this critical stochastic element in a Boltzmann machine. The dynamic cooling effect in simulated annealing can be emulated in this multi-terminal memristive device through electrostatic bias with sigmoidal thresholding distributions. We also show that a machine-learning-based method is efficient for device-circuit co-design of the Boltzmann machine based on the stochastic memristor devices in simulated annealing. The experimental demonstrations of the tunable stochastic memristors combined with the machine-learning-based device-circuit co-optimization approach for stochastic-memristor-based neural-network circuits chart a pathway for the efficient hardware realization of stochastic neural networks with applications in a broad range of electronics and computing disciplines.

cs.ET↗

arXiv2019

Multi-Interest Network with Dynamic Routing for Recommendation at Tmall

Industrial recommender systems usually consist of the matching stage and the ranking stage, in order to handle the billion-scale of users and items. The matching stage retrieves candidate items relevant to user interests, while the ranking stage sorts candidate items by user interests. Thus, the most critical ability is to model and represent user interests for either stage. Most of the existing deep learning-based models represent one user as a single vector which is insufficient to capture the varying nature of user's interests. In this paper, we approach this problem from a different view, to represent one user with multiple vectors encoding the different aspects of the user's interests. We propose the Multi-Interest Network with Dynamic routing (MIND) for dealing with user's diverse interests in the matching stage. Specifically, we design a multi-interest extractor layer based on capsule routing mechanism, which is applicable for clustering historical behaviors and extracting diverse interests. Furthermore, we develop a technique named label-aware attention to help learn a user representation with multiple vectors. Through extensive experiments on several public benchmarks and one large-scale industrial dataset from Tmall, we demonstrate that MIND can achieve superior performance than state-of-the-art methods for recommendation. Currently, MIND has been deployed for handling major online traffic at the homepage on Mobile Tmall App.

cs.IR↗

arXiv2018

Segregation assisted grain boundary precipitation in a model Al-Zn-Mg-Cu alloy

Understanding the composition evolution of grain boundaries and grain boundary precipitation at near-atomic scale in aluminum alloys is crucial to tailor mechanical properties and to increase resistance to corrosion and stress corrosion cracking. Here, we elucidate the sequence of precipitation on grain boundaries in comparison to the bulk in a model Al-Zn-Mg-Cu alloy. We investigate the material from the solution heat treated state (475{\textdegree}C), through the very early stages of aging to the peak aged state at 120{\textdegree}C and further into the overaged regime at 180{\textdegree}C. The process starts with solute enrichment on grain boundaries due to equilibrium segregation accompanied by solute depletion in their vicinity, the formation of Guinier--Preston (GP) zones in the solute-enriched grain boundary regions, and GP zones growth and transformation. The equilibrium segregation of solutes to grain boundaries during aging accelerates this sequence compared to the bulk. Analysis of the ~10 nm wide precipitate-free zones (PFZs) adjacent to the solute-enriched grain boundaries 2 shows that the depletion zones are determined by (i) interface equilibrium segregation; (ii) formation and coarsening of the grain boundary precipitates and (iii) the diffusion range of solutes in the matrix. In addition, we quantify the difference in kinetics between grain boundary and bulk precipitation. The precipitation kinetics, as observed in terms of volume fraction, average radius, and number density, is almost identical next to the depletion zone in the bulk and far inside the bulk grain remote from any grain boundary influence. This observation shows that the region influenced by the grain boundaries does not extend beyond the PFZs.

cond-mat.mtrl-sci↗

arXiv2018

Mid-wave and Long-wave IR Linear Dichroism in a Hexagonal Perovskite Chalcogenide

Mid-wave infrared (IR) and long-wave IR spectral ranges are of growing interest in various applications such as thermal imaging, thermography-based remote sensing, and night vision. Materials widely used for IR photodetectors in this regime include cadmium mercury telluride alloys and nanostructures of compound semiconductor. The materials development for IR optics will drive down the cost of IR optical systems and enable larger scale deployment. Here, we report a mid-wave IR responsive material composed of earth abundant and non-toxic elements, Sr1+xTiS3. It has a highly anisotropic quasi-one-dimensional structure similar to hexagonal perovskites. We grew large, high quality single crystals and studied its anisotropic optical properties. We observed two distinct optical absorption edges at ~2.5 um and ~5 um, respectively, for linear polarizations along two principal axes. The material demonstrated strong and broadband linear dichroism spanning mid-wave IR and long-wave IR, with a dichroitic ratio of up to 22.

cond-mat.mtrl-sci↗

arXiv2018

Parameter free quantitative analysis of atom probe data by correlation functions: Application to the precipitation in Al-Zn-Mg-Cu

Atom probe tomography enables precise quantification of the composition of second phase particles from their early stages, leading to improved understanding of the thermodynamic and kinetic mechanisms of phase formation and quantify structure-property relationships. Here we demonstrate how approaches developed for small-angle scattering can be adapted to atom probe tomography. By exploiting nearest-neighbor distributions and radial distribution function, we introduce a parameter free methodology to efficiently extract information such as particle size, composition, volume fraction, number density and inter-particle distance. We demonstrate the strength of this approach in the analysis of a precipitation-hardened model Al-Zn-Mg-Cu high-strength lightweight alloy.

cond-mat.mtrl-sci↗

arXiv2018

Billion-scale Commodity Embedding for E-commerce Recommendation in Alibaba

Recommender systems (RSs) have been the most important technology for increasing the business in Taobao, the largest online consumer-to-consumer (C2C) platform in China. The billion-scale data in Taobao creates three major challenges to Taobao's RS: scalability, sparsity and cold start. In this paper, we present our technical solutions to address these three challenges. The methods are based on the graph embedding framework. We first construct an item graph from users' behavior history. Each item is then represented as a vector using graph embedding. The item embeddings are employed to compute pairwise similarities between all items, which are then used in the recommendation process. To alleviate the sparsity and cold start problems, side information is incorporated into the embedding framework. We propose two aggregation methods to integrate the embeddings of items and the corresponding side information. Experimental results from offline experiments show that methods incorporating side information are superior to those that do not. Further, we describe the platform upon which the embedding methods are deployed and the workflow to process the billion-scale data in Taobao. Using online A/B test, we show that the online Click-Through-Rate (CTRs) are improved comparing to the previous recommendation methods widely used in Taobao, further demonstrating the effectiveness and feasibility of our proposed methods in Taobao's live production environment.

cs.IR↗

arXiv2017

Collaborative Filtering with Social Local Models

Matrix Factorization (MF) is a very popular method for recommendation systems. It assumes that the underneath rating matrix is low-rank. However, this assumption can be too restrictive to capture complex relationships and interactions among users and items. Recently, Local LOw-Rank Matrix Approximation (LLORMA) has been shown to be very successful in addressing this issue. It just assumes the rating matrix is composed of a number of low-rank submatrices constructed from subsets of similar users and items. Although LLORMA outperforms MF, how to construct such submatrices remains a big problem. Motivated by the availability of rich social connections in today's recommendation systems, we propose a novel framework, i.e., Social LOcal low-rank Matrix Approximation (SLOMA), to address this problem. To the best of our knowledge, SLOMA is the first work to incorporate social connections into the local low-rank framework. Furthermore, we enhance SLOMA by applying social regularization to submatrices factorization, denoted as SLOMA++. Therefore, the proposed model can benefit from both social recommendation and the local low-rank assumption. Experimental results from two real-world datasets, Yelp and Douban, demonstrate the superiority of the proposed models over LLORMA and MF.

cs.SI↗

arXiv2017

Atomically-thin Femtojoule Filamentary Memristor

The morphology and dimension of the conductive filament formed in a memristive device are strongly influenced by the thickness of its switching medium layer. Aggressive scaling of this active layer thickness is critical towards reducing the operating current, voltage and energy consumption in filamentary type memristors. Previously, the thickness of this filament layer has been limited to above a few nanometers due to processing constraints, making it challenging to further suppress the on-state current and the switching voltage. Here, we study the formation of conductive filaments in a material medium with sub-nanometer thickness, formed through the oxidation of atomically-thin two-dimensional boron nitride. The resulting memristive device exhibits sub-nanometer filamentary switching with sub-pA operation current and femtojoule per bit energy consumption. Furthermore, by confining the filament to the atomic scale, we observe current switching characteristics that are distinct from that in thicker medium due to the profoundly different atomic kinetics. The filament morphology in such an aggressively scaled memristive device is also theoretically explored. These ultra-low energy devices are promising for realizing femtojoule and sub-femtojoule electronic computation, which can be attractive for applications in a wide range of electronics systems that desire ultra-low power operation.

physics.app-ph↗

arXiv2017

Spatial-Temporal Imaging of Anisotropic Photocarrier Dynamics in Black Phosphorus

As an emerging single elemental layered material with a low symmetry in-plane crystal lattice, black phosphorus (BP) has attracted significant research interest owing to its unique electronic and optoelectronic properties, including its widely tunable bandgap, polarization dependent photoresponse and highly anisotropic in-plane charge transport. Despite extensive study of the steady-state charge transport in BP, there has not been direct characterization and visualization of the hot carriers dynamics in BP immediately after photoexcitation, which is crucial to understanding the performance of BP-based optoelectronic devices. Here we use the newly developed scanning ultrafast electron microscopy (SUEM) to directly visualize the motion of photo-excited hot carriers on the surface of BP in both space and time. We observe highly anisotropic in-plane diffusion of hot holes, with a 15-times higher diffusivity along the armchair (x-) direction than that along the zigzag (y-) direction. Our results provide direct evidence of anisotropic hot carrier transport in BP and demonstrate the capability of SUEM to resolve ultrafast hot carrier dynamics in layered two-dimensional materials.

cond-mat.mtrl-sci↗

arXiv2016

Anisotropic Black Phosphorus Synaptic Device for Neuromorphic Applications

Synapses are functional links between neurons, through which "information" flows in the neural network. These connections vary significantly in strength, typically resulting from the intrinsic heterogeneity in their chemical and biological properties. Such heterogeneity is fundamental to the diversity of neural activities, which together with other features of the brain enables functions ranging from perception and recognition, to memory and reasoning. Realizing such heterogeneity in synaptic electronics is critical towards building artificial neural network with the potential for achieving the level of complexity in biological systems. However, such intrinsic heterogeneity has been very challenging to realize in current synaptic devices. Here, we demonstrate the first black phosphorus (BP) synaptic device, which offers intrinsic anisotropy in its synaptic characteristics directly resulting from its low crystal symmetry. The charge transfer between the 2-nm native oxide of BP and the BP channel is utilized to achieve the synaptic behavior. Key features of biological synapses such as long-term plasticity with heterogeneity, including long-term potentiation/depression and spike-timing-dependent plasticity, are mimicked. With the anisotropic BP synaptic devices, we also realize a simple compact heterogeneous axon-multi-synapses network. This demonstration represents an important step towards introducing intrinsic heterogeneity to artificial neuromorphic systems.

cond-mat.mtrl-sci↗

arXiv2015

Monolayer Molybdenum Disulfide Nanoribbons with High Optical Anisotropy

Two-dimensional Molybdenum Disulfide (MoS2) has shown promising prospects for the next generation electronics and optoelectronics devices. The monolayer MoS2 can be patterned into quasi-one-dimensional anisotropic MoS2 nanoribbons (MNRs), in which theoretical calculations have predicted novel properties. However, little work has been carried out in the experimental exploration of MNRs with a width of less than 20 nm where the geometrical confinement can lead to interesting phenomenon. Here, we prepared MNRs with width between 5 nm to 15 nm by direct helium ion beam milling. High optical anisotropy of these MNRs is revealed by the systematic study of optical contrast and Raman spectroscopy. The Raman modes in MNRs show strong polarization dependence. Besides that the E' and A'1 peaks are broadened by the phonon-confinement effect, the modes corresponding to singularities of vibrational density of states are activated by edges. The peculiar polarization behavior of Raman modes can be explained by the anisotropy of light absorption in MNRs, which is evidenced by the polarized optical contrast. The study opens the possibility to explore quasione-dimensional materials with high optical anisotropy from isotropic 2D family of transition metal dichalcogenides.

cond-mat.mtrl-sci↗

arXiv2015

Interlayer Interactions in Anisotropic Atomically-thin Rhenium Diselenide

Recently, two-dimensional (2D) materials with strong in-plane anisotropic properties such as black phosphorus have demonstrated great potential for developing new devices that can take advantage of its reduced lattice symmetry with potential applications in electronics, optoelectronics and thermoelectrics. However, the selection of 2D material with strong in-plane anisotropy has so far been very limited and only sporadic studies have been devoted to transition metal dichalcogenides (TMDC) materials with reduced lattice symmetry, which is yet to convey the full picture of their optical and phonon properties, and the anisotropy in their interlayer interactions. Here, we study the anisotropic interlayer interactions in an important TMDC 2D material with reduced in-plane symmetry - atomically thin rhenium diselenide (ReSe2) - by investigating its ultralow frequency interlayer phonon vibration modes, the layer dependent optical bandgap, and the anisotropic photoluminescence (PL) spectra for the first time. The ultralow frequency interlayer Raman spectra combined with the first study of polarization-resolved high frequency Raman spectra in mono- and bi-layer ReSe2 allows deterministic identification of its layer number and crystal orientation. PL measurements show anisotropic optical emission intensity with bandgap increasing from 1.26 eV in the bulk to 1.32 eV in monolayer, consistent with the theoretical results based on first-principle calculations. The study of the layer-number dependence of the Raman modes and the PL spectra reveals the relatively weak van der Waals interaction and 2D quantum confinement in atomically-thin ReSe2.

cond-mat.mes-hall↗

arXiv2014

Highly Anisotropic and Robust Excitons in Monolayer Black Phosphorus

Semi-metallic graphene and semiconducting monolayer transition metal dichalcogenides (TMDCs) are the two-dimensional (2D) materials most intensively studied in recent years. Recently, black phosphorus emerged as a promising new 2D material due to its widely tunable and direct bandgap, high carrier mobility and remarkable in-plane anisotropic electrical, optical and phonon properties. However, current progress is primarily limited to its thin-film form, and its unique properties at the truly 2D quantum confinement have yet to be demonstrated. Here, we reveal highly anisotropic and tightly bound excitons in monolayer black phosphorus using polarization-resolved photoluminescence measurements at room temperature. We show that regardless of the excitation laser polarization, the emitted light from the monolayer is linearly polarized along the light effective mass direction and centers around 1.3 eV, a clear signature of emission from highly anisotropic bright excitons. In addition, photoluminescence excitation spectroscopy suggests a quasiparticle bandgap of 2.2 eV, from which we estimate an exciton binding energy of around 0.9 eV, consistent with theoretical results based on first-principles. The experimental observation of highly anisotropic, bright excitons with exceedingly large binding energy not only opens avenues for the future explorations of many-electron effects in this unusual 2D material, but also suggests a promising future in optoelectronic devices such as on-chip infrared light sources.

cond-mat.mes-hall↗