arXiv subjects

Jie Ren

Publications and source records attributed to Jie Ren.

At least 253 records · Page 14Linked to original sources

arXiv2019

SummAE: Zero-Shot Abstractive Text Summarization using Length-Agnostic Auto-Encoders

We propose an end-to-end neural model for zero-shot abstractive text summarization of paragraphs, and introduce a benchmark task, ROCSumm, based on ROCStories, a subset for which we collected human summaries. In this task, five-sentence stories (paragraphs) are summarized with one sentence, using human summaries only for evaluation. We show results for extractive and human baselines to demonstrate a large abstractive gap in performance. Our model, SummAE, consists of a denoising auto-encoder that embeds sentences and paragraphs in a common space, from which either can be decoded. Summaries for paragraphs are generated by decoding a sentence from the paragraph representations. We find that traditional sequence-to-sequence auto-encoders fail to produce good summaries and describe how specific architectural choices and pre-training techniques can significantly improve performance, outperforming extractive baselines. The data, training, evaluation code, and best model weights are open-sourced.

cs.CL↗

arXiv2019

Semi-parametric Bayesian variable selection for gene-environment interactions

Many complex diseases are known to be affected by the interactions between genetic variants and environmental exposures beyond the main genetic and environmental effects. Study of gene-environment (G$\times$E) interactions is important for elucidating the disease etiology. Existing Bayesian methods for G$\times$E interaction studies are challenged by the high-dimensional nature of the study and the complexity of environmental influences. Many studies have shown the advantages of penalization methods in detecting G$\times$E interactions in "large p, small n" settings. However, Bayesian variable selection, which can provide fresh insight into G$\times$E study, has not been widely examined. We propose a novel and powerful semi-parametric Bayesian variable selection model that can investigate linear and nonlinear G$\times$E interactions simultaneously. Furthermore, the proposed method can conduct structural identification by distinguishing nonlinear interactions from main-effects-only case within the Bayesian framework. Spike and slab priors are incorporated on both individual and group levels to identify the sparse main and interaction effects. The proposed method conducts Bayesian variable selection more efficiently than existing methods. Simulation shows that the proposed model outperforms competing alternatives in terms of both identification and prediction. The proposed Bayesian method leads to the identification of main and interaction effects with important implications in a high-throughput profiling study with high-dimensional SNP data.

stat.ME↗

arXiv2019

Spectral Weight Suppression and Fermi Arc-like Features with Strong Holographic Lattices

We investigate holographic fermions in uni-directional striped phases, where the breaking of translational invariance can be generated either spontaneously or explicitly. We solve the Dirac equation for a probe fermion in the associated background geometry. When the spatial modulation effect becomes sufficiently strong, we see a spectral weight suppression whenever the Fermi surface is larger than the first Brillouin zone. This leads to the gradual disappearance of the Fermi surface along the symmetry breaking direction, in all of the cases we have examined. This effect appears to be a generic consequence of strong inhomogeneities, independently of whether translational invariance is broken spontaneously or explicitly. The resulting Fermi surface is segmented and has features reminiscent of Fermi arcs.

hep-th↗

arXiv2019

Sentinel: Runtime Data Management on Heterogeneous Main MemorySystems for Deep Learning

Software-managed heterogeneous memory (HM) provides a promising solution to increase memory capacity and cost efficiency. However, to release the performance potential of HM, we face a problem of data management. Given an application with various execution phases and each with possibly distinct working sets, we must move data between memory components of HM to optimize performance. The deep neural network (DNN), as a common workload on data centers, imposes great challenges on data management on HM. This workload often employs a task dataflow execution model, and is featured with a large amount of small data objects and fine-grained operations (tasks). This execution model imposes challenges on memory profiling and efficient data migration. We present Sentinel, a runtime system that automatically optimizes data migration (i.e., data management) on HM to achieve performance similar to that on the fast memory-only system with a much smaller capacity of fast memory. To achieve this,Sentinel exploits domain knowledge about deep learning to adopt a custom approach for data management. Sentinel leverages workload repeatability to break the dilemma between profiling accuracy and overhead; It enables profiling and data migration at the granularity of data objects (not pages), by controlling memory allocation. This method bridges the semantic gap between operating system and applications. By associating data objects with the DNN topology, Sentinel avoids unnecessary data movement and proactively triggers data movement. Using only 20% of peak memory consumption of DNN models as fast memory size, Sentinel achieves the same or comparable performance (at most 8% performance difference) to that of the fast memory-only system on common DNN models; Sentinel also consistently outperforms a state-of-the-art solution by 18%.

cs.PF↗

arXiv2019

Using Machine Learning to Optimize Web Interactions on Heterogeneous Mobile Multi-cores

The web has become a ubiquitous application development platform for mobile systems. Yet, web access on mobile devices remains an energy-hungry activity. Prior work in the field mainly focuses on the initial page loading stage, but fails to exploit the opportunities for energy-efficiency optimization while the user is interacting with a loaded page. This paper presents a novel approach for performing energy optimization for interactive mobile web browsing. At the heart of our approach is a set of machine learning models, which estimate \emph{at runtime} the frames per second for a given user interaction input by running the computation-intensive web render engine on a specific processor core under a given clock speed. We use the learned predictive models as a utility function to quickly search for the optimal processor setting to carefully trade responsive time for reduced energy consumption. We integrate our techniques to the open-source Chromium browser and apply it to two representative mobile user events: scrolling and pinching (i.e., zoom in and out). We evaluate the developed system on the landing pages of the top-100 hottest websites and two big.LITTLE heterogeneous mobile platforms. Our extensive experiments show that the proposed approach reduces the system-wide energy consumption by over 36\% on average and up to 70\%. This translates to an over 17\% improvement on energy-efficiency over a state-of-the-art event-based web browser scheduler, but with significantly fewer violations on the quality of service.

cs.DC↗

arXiv2019

Probing stop pair production at the LHC with graph neural networks

Top-squarks (stops) play a crucial role for the naturalness of supersymmetry (SUSY). However, searching for the stops is a tough task at the LHC. To dig the stops out of the huge LHC data, various expert-constructed kinematic variables or cutting-edge analysis techniques have been invented. In this paper, we propose to represent collision events as event graphs and use the message passing neutral network (MPNN) to analyze the events. As a proof-of-concept, we use our method in the search of the stop pair production at the LHC, and find that our MPNN can efficiently discriminate the signal and background events. In comparison with other machine learning methods (e.g. DNN), MPNN can enhance the mass reach of stop mass by several tens of GeV to over a hundred GeV.

hep-ph↗

arXiv2019

Supervised deep learning in high energy phenomenology: a mini review

Deep learning, a branch of machine learning, have been recently applied to high energy experimental and phenomenological studies. In this note we give a brief review on those applications using supervised deep learning. We first describe various learning models and then recapitulate their applications to high energy phenomenological studies. Some detailed applications are delineated in details, including the machine learning scan in the analysis of new physics parameter space, the graph neural networks in the search of top-squark production and in the $CP$ measurement of the top-Higgs coupling at the LHC.

hep-ph↗

arXiv2019

EasyCrash: Exploring Non-Volatility of Non-Volatile Memory for High Performance Computing Under Failures

Emerging non-volatile memory (NVM) is promising for building future HPC. Leveraging the non-volatility of NVM as main memory, we can restart the application using data objects remaining on NVM when the application crashes. This paper explores this solution to handle HPC under failures, based on the observation that many HPC applications have good enough intrinsic fault tolerance. To improve the possibility of successful recomputation with correct outcomes and ignorable performance loss, we introduce EasyCrash, a framework to decide how to selectively persist application data objects during application execution. Our evaluation shows that EasyCrash transforms 54% of crashes that cannot correctly recompute into the correct computation while incurring a negligible performance overhead (1.5% on average). Using EasyCrash and application intrinsic fault tolerance, 82% of crashes can successfully recompute. When EasyCrash is used with a traditional checkpoint scheme, it enables up to 24% improvement (15% on average) in system efficiency.

cs.PF↗

arXiv2019

High Thermoelectric Performance in Two-Dimensional Tellurium: An Ab Initio Study

In 2016, bulk tellurium was experimentally observed as a remarkable thermoelectric material. Recently, two-dimensional (2D) tellurium, called tellurene, has been synthesized and has exhibited unexpected electronic properties compared with the 2D MoS$_2$. They have also been fabricated into air-stable and high efficient field-effect transistors. There are two stable 2D tellurene phases. One ($β$-Te) has been confirmed with an ultralow lattice thermal conductivity ($κ_L$). However, the study of the transport properties of the other more stable phase, $α$-Te, is still lacking. Here, we report the thermoelectric performance and phonon properties of $α$-Te using Boltzmann transport theory and first principle calculations. A maximum ZT value of 0.83 is achieved under reasonable hole concentration, suggesting that the monolayer $α$-Te is a potential competitor in the thermoelectric field.

physics.comp-ph↗

arXiv2019

Energy Super-Diffusion in One-Dimensional Momentum Non-Conserving Nonlinear Lattices

There is a well-known mapping between energy normal (super-) diffusion and normal (anomalous) heat conduction in one-dimensional (1D) nonlinear lattices. The momentum conserving nonlinear lattices exhibit energy super-diffusion behavior with the only exception of coupled rotator model. Yet, for all other 1D momentum nonconserving nonlinear lattices studied so far, the energy diffusion or heat conduction is normal. Here we propose a 1D nonlinear lattice model with negative couplings, which is momentum non-conserving due to the translational symmetry breaking. Our numerical results show that energy super-diffusion instead of normal diffusion can be found for this model, which indicates that neither momentum non-conservation is a sufficient condition for energy normal diffusion nor momentum conservation is a necessary condition for energy super-diffusion. Zero frequency phonon mode at Brillouin zone boundary induces a new conserved momentum parity, which is the key for the energy super-diffusion and anomalous heat conduction. Removing the zero frequency mode, such as by on-site potential, is a sufficient condition for normal heat conduction in 1D nonlinear lattices.

cond-mat.stat-mech↗

arXiv2019

Anisotropic thermal expansion and thermomechanic properties of monolayer $β$-Te

Recently, $β$-Te (atomically 2D tellurium) with rectangular crystal structure has been synthesized successfully on highly oriented pyrolytic graphite substrates by using molecular beam epitaxy. It has been found possessing remarkable properties such as ultralow lattice thermal conductivity and high thermoelectric efficiency. Based on the first-principles calculations, we study the thermal expansion and thermomechanic properties of the experimental phase monolayer $β$-Te, using quasiharmonic approach. It is found $β$-Te shows large positive thermal expansion at elevated temperature, while the linear thermal expansion coefficient is negative along a direction at very low temperature. The linear thermal expansion coefficient along b direction is 4.9*10$^{-5}$ K$^{-1}$ at 500 K, which is considerably large in 2D materials. $β$-Te exhibits strong in-plane anisotropy, including thermal expansion, 2D elastic moduli and Poisson's ratios. However, the elastic moduli, Poisson's ratios and the in-plane anisotropy are weakened with increasing temperature, and the variations are dominated by the generalized mode Grüneisen parameters.

cond-mat.mtrl-sci↗

arXiv2019

Architecture-Aware, High Performance Transaction for Persistent Memory

Byte-addressable non-volatile main memory (NVM) demands transactional mechanisms to access and manipulate data on NVM atomically. Those transaction mechanisms often employ a logging mechanism (undo logging or redo logging). However, the logging mechanisms can bring large runtime overhead (8%-49% in our evaluation), and 41%-78% of the overhead attributes to the frequent cache-line flushing. Such large overhead significantly diminishes the performance benefits offered by NVM. In this paper, we introduce a new method to reduce the overhead of cache-line flushing for logging-based transactions. Different from the traditional method that works at the program level and leverages program semantics to reduce the logging overhead, we introduce architecture awareness. In particular, we do not flush certain cache blocks, as long as they are estimated to be eliminated out of the cache because of the hardware caching mechanism (e.g., the cache replacement algorithm). Furthermore, we coalesce those cache blocks with low dirtiness to improve the efficiency of cache-line flushing. We implement an architecture-aware, high performance transaction runtime system for persistent memory, Archapt. Our results show that comparing with an undo logging (PMDK) and a redo logging (Mnemosyne), Archapt reduces cache-line flushing by 66% and improves system throughput by 19% on average (42% at most). Our crash tests with four hardware caching policies show that Archapt provides a strong guarantee on crash consistency.

cs.DC↗

arXiv2019

Zero-index and Topology in 1D Phononic Metamaterials with Negative Mass and Negative Coupling

Phononic metamaterials have attracted extensive attention since they are exibly adjustable to control the phonon transmission. In this work, we study a one-dimensional phononic metamaterial, made of mechanical resonant oscillators and chiral couplings. We show that by design, the oscillator mass and inter-oscillator coupling, although both are positive naturally, can be either single negative or double negative effectively within a certain frequency range. At the frequency where the effective mass and coupling are both infinite, a flat band emerges that will induce an extremely high density of states. At the critical point of band degeneracy, a Dirac-like point emerges where both effective mass and the inverse of effective coupling are simultaneously zero, so that zero index is realized for phonons. Moreover, the phononic topological phase transition is observed that the phononic band gap switches between single mass-negative and single coupling-negative regime. As a consequence, a topological interface state is identified, well explained by the theory.

cond-mat.mes-hall↗

arXiv2019

Exploring supersymmetry with machine learning

Investigation of well-motivated parameter space in the theories of Beyond the Standard Model (BSM) plays an important role in new physics discoveries. However, a large-scale exploration of models with multi-parameter or equivalent solutions with a finite separation, such as supersymmetric models, is typically a time-consuming and challenging task. In this paper, we propose a self-exploration method, named Machine Learning Scan (MLS), to achieve an efficient test of models. As a proof-of-concept, we apply MLS to investigate the subspace of MSSM and CMSSM and find that such a method can reduce the computational cost and may be helpful for accelerating the exploration of supersymmetry.

hep-ph↗

arXiv2019

Strong system-bath coupling induces negative differential thermal conductance and heat amplification in nonequilibrium two-qubits systems

Quantum heat transfer is analyzed in nonequilibrium two-qubits systems by applying the nonequilibrium polaron-transformed Redfield equation combined with full counting statistics. Steady state heat currents with weak and strong qubit-bath couplings are clearly unified. Within the two-terminal setup, the negative differential thermal conductance is unraveled with strong qubit-bath coupling and finite qubit splitting energy. The partially strong spin-boson interaction is sufficient to show the negative differential thermal conductance. Based on the three-terminal setup, that two-qubits are asymmetrically coupled to three thermal baths, a giant heat amplification factor is observed with strong qubit-bath coupling. Moreover, the strong interaction of either the left or right spin-boson coupling is able to exhibit the apparent heat amplification effect.

quant-ph↗

arXiv2019

Unveiling CP property of top-Higgs coupling with graph neural networks at the LHC

The top-Higgs coupling plays an important role in particle physics and cosmology. The precision measurements of this coupling can provide an insight to new physics beyond the Standard Model. In this paper, we propose to use Message Passing Neural Network (MPNN) to reveal the CP nature of top-Higgs interaction through semi-leptonic channel $pp \to t(\to b\ell^-ν_\ell)\bar{t}(\to \bar{b}jj)h(\to b\bar{b})$. Using the test statistics constructed from the event classification probabilities given by the MPNN, we find that the pure CP-even and CP-odd components can be well distinguished at the LHC, with at most 300 fb$^{-1}$ experimental data.

hep-ph↗

arXiv2019

Anomalous near-field heat transfer in carbon-based nano-structures with edge states

We find an unusually optimal near field heat transfer, where the maximum heat transfer is reached at experimentally feasible gap separation. We attribute this to the localized zero-energy electronic edge states, which also substantially changes the near-field behaviors. We demonstrate these anomalous behaviors in two typical carbon-based nano-structures: zigzag single-walled carbon nanotubes and graphene nano-triangles. For the system of carbon nanotubes, the maximal heat flux in this work surpasses all the previous results reported so far by several orders of magnitude. The underlying mechanisms for the peculiar effects are uncovered from a simple Su-Schrieffer-Heeger model. Our findings also offer a novel route to active near-field thermal switch, where the heat flux can be modulated through tuning the presence or absence of edge states.

cond-mat.mes-hall↗

arXiv2018

Phase diagram of the spin-1/2 Heisenberg alternating chain in a magnetic field

By using the infinite time-evolving block decimation, we study quantum fidelity and entanglement entropy in the spin-1/2 Heisenberg alternating chain under an external magnetic field. The effects of the magnetic field on the fidelity are investigated, and its relation with the quantum phase transition (QPT) is analyzed. The phase diagram of the model is given accordingly, which supports the Haldane phase, the singlet-dimer phase, the Luttinger liquid phase and the paramagnetic phase. The scaling of entanglement entropy in the gapless Luttinger liquid phase is studied, and the central charge c = 1 is obtained. We also study the relationship between the quantum coherence, string order parameter and QPTs. Results obtained from these quantum information observations are consistent with the previous reports.

quant-ph↗