Showing 1–28 of 28 results for author: Che, S

Data Caching for Enterprise-Grade Petabyte-Scale OLAP

Authors: Chunxu Tang, Bin Fan, Jing Zhao, Chen Liang, Yi Wang, Beinan Wang, Ziyue Qiu, Lu Qiu, Bowen Ding, Shouzhuo Sun, Saiguang Che, Jiaming Mai, Shouwei Chen, Yu Zhu, Jianjian Xie, Yutian, Sun, Yao Li, Yangjun Zhang, Ke Wang, Mingmin Chen

Abstract: With the exponential growth of data and evolving use cases, petabyte-scale OLAP data platforms are increasingly adopting a model that decouples compute from storage. This shift, evident in organizations like Uber and Meta, introduces operational challenges including massive, read-heavy I/O traffic with potential throttling, as well as skewed and fragmented data access patterns. Addressing these ch… ▽ More With the exponential growth of data and evolving use cases, petabyte-scale OLAP data platforms are increasingly adopting a model that decouples compute from storage. This shift, evident in organizations like Uber and Meta, introduces operational challenges including massive, read-heavy I/O traffic with potential throttling, as well as skewed and fragmented data access patterns. Addressing these challenges, this paper introduces the Alluxio local (edge) cache, a highly effective architectural optimization tailored for such environments. This embeddable cache, optimized for petabyte-scale data analytics, leverages local SSD resources to alleviate network I/O and API call pressures, significantly improving data transfer efficiency. Integrated with OLAP systems like Presto and storage services like HDFS, the Alluxio local cache has demonstrated its effectiveness in handling large-scale, enterprise-grade workloads over three years of deployment at Uber and Meta. We share insights and operational experiences in implementing these optimizations, providing valuable perspectives on managing modern, massive-scale OLAP workloads. △ Less

Submitted 9 June, 2024; originally announced June 2024.

Comments: Accepted to the USENIX Annual Technical Conference (USENIX ATC) 2024

Fermihedral: On the Optimal Compilation for Fermion-to-Qubit Encoding

Authors: Yuhao Liu, Shize Che, Junyu Zhou, Yunong Shi, Gushu Li

Abstract: This paper introduces Fermihedral, a compiler framework focusing on discovering the optimal Fermion-to-qubit encoding for targeted Fermionic Hamiltonians. Fermion-to-qubit encoding is a crucial step in harnessing quantum computing for efficient simulation of Fermionic quantum systems. Utilizing Pauli algebra, Fermihedral redefines complex constraints and objectives of Fermion-to-qubit encoding int… ▽ More This paper introduces Fermihedral, a compiler framework focusing on discovering the optimal Fermion-to-qubit encoding for targeted Fermionic Hamiltonians. Fermion-to-qubit encoding is a crucial step in harnessing quantum computing for efficient simulation of Fermionic quantum systems. Utilizing Pauli algebra, Fermihedral redefines complex constraints and objectives of Fermion-to-qubit encoding into a Boolean Satisfiability problem which can then be solved with high-performance solvers. To accommodate larger-scale scenarios, this paper proposed two new strategies that yield approximate optimal solutions mitigating the overhead from the exponentially large number of clauses. Evaluation across diverse Fermionic systems highlights the superiority of Fermihedral, showcasing substantial reductions in implementation costs, gate counts, and circuit depth in the compiled circuits. Real-system experiments on IonQ's device affirm its effectiveness, notably enhancing simulation accuracy. △ Less

Submitted 26 March, 2024; v1 submitted 26 March, 2024; originally announced March 2024.

Journal ref: ASPLOS 2024

Dominant 1/3-filling Correlated Insulator States and Orbital Geometric Frustration in Twisted Bilayer Graphene

Authors: Haidong Tian, Emilio Codecido, Dan Mao, Kevin Zhang, Shi Che, Kenji Watanabe, Takashi Taniguchi, Dmitry Smirnov, Eun-Ah Kim, Marc Bockrath, Chun Ning Lau

Abstract: Geometric frustration is a phenomenon in a lattice system where not all interactions can be satisfied, the simplest example being antiferromagnetically coupled spins on a triangular lattice. Frustrated systems are characterized by their many nearly degenerate ground states, leading to non-trivial phases such as spin ice and spin liquids. To date most studies are on geometric frustration of spins;… ▽ More Geometric frustration is a phenomenon in a lattice system where not all interactions can be satisfied, the simplest example being antiferromagnetically coupled spins on a triangular lattice. Frustrated systems are characterized by their many nearly degenerate ground states, leading to non-trivial phases such as spin ice and spin liquids. To date most studies are on geometric frustration of spins; much less explored is orbital geometric frustration. For electrons in twisted bilayer graphene (tBLG) at denominator 3 fractional filling, Coulomb interactions and the Wannier orbital shapes are predicted to strongly constrain spatial charge ordering, leading to geometrically frustrated ground states that produce a new class of correlated insulators (CIs). Here we report the observation of dominant denominator 3 fractional filling insulating states in large angle tBLG; these states persist in magnetic fields and display magnetic ordering signatures and tripled unit cell reconstruction. These results are in agreement with a strong-coupling theory of symmetry-breaking of geometrically frustrated fractional states. △ Less

Submitted 22 February, 2024; originally announced February 2024.

Bias of AI-Generated Content: An Examination of News Produced by Large Language Models

Authors: Xiao Fang, Shangkun Che, Minjia Mao, Hongzhe Zhang, Ming Zhao, Xiaohang Zhao

Abstract: Large language models (LLMs) have the potential to transform our lives and work through the content they generate, known as AI-Generated Content (AIGC). To harness this transformation, we need to understand the limitations of LLMs. Here, we investigate the bias of AIGC produced by seven representative LLMs, including ChatGPT and LLaMA. We collect news articles from The New York Times and Reuters,… ▽ More Large language models (LLMs) have the potential to transform our lives and work through the content they generate, known as AI-Generated Content (AIGC). To harness this transformation, we need to understand the limitations of LLMs. Here, we investigate the bias of AIGC produced by seven representative LLMs, including ChatGPT and LLaMA. We collect news articles from The New York Times and Reuters, both known for their dedication to provide unbiased news. We then apply each examined LLM to generate news content with headlines of these news articles as prompts, and evaluate the gender and racial biases of the AIGC produced by the LLM by comparing the AIGC and the original news articles. We further analyze the gender bias of each LLM under biased prompts by adding gender-biased messages to prompts constructed from these news headlines. Our study reveals that the AIGC produced by each examined LLM demonstrates substantial gender and racial biases. Moreover, the AIGC generated by each LLM exhibits notable discrimination against females and individuals of the Black race. Among the LLMs, the AIGC generated by ChatGPT demonstrates the lowest level of bias, and ChatGPT is the sole model capable of declining content generation when provided with biased prompts. △ Less

Submitted 3 April, 2024; v1 submitted 18 September, 2023; originally announced September 2023.

DeepSpeed-Chat: Easy, Fast and Affordable RLHF Training of ChatGPT-like Models at All Scales

Authors: Zhewei Yao, Reza Yazdani Aminabadi, Olatunji Ruwase, Samyam Rajbhandari, Xiaoxia Wu, Ammar Ahmad Awan, Jeff Rasley, Minjia Zhang, Conglong Li, Connor Holmes, Zhongzhu Zhou, Michael Wyatt, Molly Smith, Lev Kurilenko, Heyang Qin, Masahiro Tanaka, Shuai Che, Shuaiwen Leon Song, Yuxiong He

Abstract: ChatGPT-like models have revolutionized various applications in artificial intelligence, from summarization and coding to translation, matching or even surpassing human performance. However, the current landscape lacks an accessible, efficient, and cost-effective end-to-end RLHF (Reinforcement Learning with Human Feedback) training pipeline for these powerful models, particularly when training at… ▽ More ChatGPT-like models have revolutionized various applications in artificial intelligence, from summarization and coding to translation, matching or even surpassing human performance. However, the current landscape lacks an accessible, efficient, and cost-effective end-to-end RLHF (Reinforcement Learning with Human Feedback) training pipeline for these powerful models, particularly when training at the scale of billions of parameters. This paper introduces DeepSpeed-Chat, a novel system that democratizes RLHF training, making it accessible to the AI community. DeepSpeed-Chat offers three key capabilities: an easy-to-use training and inference experience for ChatGPT-like models, a DeepSpeed-RLHF pipeline that replicates the training pipeline from InstructGPT, and a robust DeepSpeed-RLHF system that combines various optimizations for training and inference in a unified way. The system delivers unparalleled efficiency and scalability, enabling training of models with hundreds of billions of parameters in record time and at a fraction of the cost. With this development, DeepSpeed-Chat paves the way for broader access to advanced RLHF training, even for data scientists with limited resources, thereby fostering innovation and further development in the field of AI. △ Less

Submitted 2 August, 2023; originally announced August 2023.

Comments: 14 pages, 7 figures

Single Diamond Structured Titania Scaffold

Authors: Chao Wang, Congcong Cui, Quanzheng Deng, Chong Zhang, Shunsuke Asahina, Yuanyuan Cao, Yiyong Mai, Shunai Che, Lu Han

Abstract: The single diamond (SD) network, discovered in beetle and weevil skeletons, is the 'holy grail' of photonic materials with the widest complete bandgap known to date. However, the thermodynamic instability of SD has made its self-assembly long been a formidable challenge. By imitating the simultaneous co-folding process of nonequilibrium skeleton formation in natural organisms, we devised an unprec… ▽ More The single diamond (SD) network, discovered in beetle and weevil skeletons, is the 'holy grail' of photonic materials with the widest complete bandgap known to date. However, the thermodynamic instability of SD has made its self-assembly long been a formidable challenge. By imitating the simultaneous co-folding process of nonequilibrium skeleton formation in natural organisms, we devised an unprecedented bottom-up approach to fabricate SD networks via the synergistic self-assembly of diblock copolymer and inorganic precursors and successfully obtained tetrahedral connected polycrystalline anatase SD frameworks. A photonic bandstructure calculation showed that the resulting SD structure has a wide and complete photonic bandgap. This work provides an ingenious design solution to the complex synthetic puzzle and offers new opportunities for biorelevant materials, next-generation optical devices, etc. △ Less

Submitted 26 July, 2023; v1 submitted 29 June, 2023; originally announced June 2023.

Tuning Spin Transport in a Graphene Antiferromagnetic Insulator

Authors: Petr Stepanov, Dmitry L. Shcherbakov, Shi Che, Marc W. Bockrath, Yafis Barlas, Dmitry Smirnov, Kenji Watanabe, Takashi Taniguchi, Roger K. Lake, Chun Ning Lau

Abstract: Long-distance spin transport through anti-ferromagnetic insulators (AFMIs) is a long-standing goal of spintronics research. Unlike conventional spintronics systems, monolayer graphene in quantum Hall regime (QH) offers an unprecedented tuneability of spin-polarization and charge carrier density in QH edge states. Here, using gate-controlled QH edges as spin-dependent injectors and detectors in an… ▽ More Long-distance spin transport through anti-ferromagnetic insulators (AFMIs) is a long-standing goal of spintronics research. Unlike conventional spintronics systems, monolayer graphene in quantum Hall regime (QH) offers an unprecedented tuneability of spin-polarization and charge carrier density in QH edge states. Here, using gate-controlled QH edges as spin-dependent injectors and detectors in an all-graphene electrical circuit, for the first time we demonstrate a selective tuning of ambipolar spin transport through graphene $νにゅー$=0 AFMIs. By modulating polarities of the excitation bias, magnetic fields, and charge carriers that host opposite chiralities, we show that the difference between spin chemical potentials of adjacent edge channels in the spin-injector region is crucial in tuning spin-transport observed across graphene AFMI. We demonstrate that non-local response vanishes upon reversing directions of the co-propagating edge channels when the spin-filters in our devices are no longer selective for a particular spin-polarization. Our results establish a versatile set of methods to tune pure spin transport via an anti-ferromagnetic media and open a pathway to explore their applications for a broad field of antiferromagnetic spintronics research. △ Less

Submitted 25 May, 2022; originally announced May 2022.

Journal ref: Phys. Rev. Applied 18, 014031, 2022

Evidence for Flat Band Dirac Superconductor Originating from Quantum Geometry

Authors: Haidong Tian, Shi Che, Tianyi Xu, Patrick Cheung, Kenji Watanabe, Takashi Taniguchi, Mohit Randeria, Fan Zhang, Chun Ning Lau, Marc W. Bockrath

Abstract: In a flat band superconductor, the charge carriers' group velocity vF is extremely slow, quenching their kinetic energy. The emergence of superconductivity thus appears paradoxical, as conventional BCS theory implies a vanishing coherence length, superfluid stiffness, and critical current. Here, using twisted bilayer graphene (tBLG), we explore the profound effect of vanishingly small vF in a Dira… ▽ More In a flat band superconductor, the charge carriers' group velocity vF is extremely slow, quenching their kinetic energy. The emergence of superconductivity thus appears paradoxical, as conventional BCS theory implies a vanishing coherence length, superfluid stiffness, and critical current. Here, using twisted bilayer graphene (tBLG), we explore the profound effect of vanishingly small vF in a Dirac superconducting flat band system Using Schwinger-limited non-linear transport studies, we demonstrate an extremely slow vF ~ 1000 m/s for filling fraction nu between -1/2 and -3/4 of the moire superlattice. In the superconducting state, the same velocity limit constitutes a new limiting mechanism for the critical current, analogous to a relativistic superfluid. Importantly, our measurement of superfluid stiffness, which controls the superconductor's electrodynamic response, shows that it is not dominated by the kinetic energy, but instead by the interaction-driven superconducting gap, consistent with recent theories on a quantum geometric contribution. We find evidence for small pairs, characteristic of the BCS to Bose-Einstein condensation (BEC) crossover, with an unprecedented ratio of the superconducting transition temperature to the Fermi temperature exceeding unity, and discuss how this arises for very strong coupling superconductivity in ultra-flat Dirac bands. △ Less

Submitted 26 December, 2021; originally announced December 2021.

Comments: 4 figures

Hourly Warning for Strong Earthquakes

Authors: T. Chen, L. Li, X. -X. Zhang, C. Wang, X. -B. Jin, Q. -M. Ma, J. -Y. Xu, Z. -H. He, H. Li, S. -G. Xiao, X. -Z. Wang, X. -H. Shen, X. -M. Zhang, H. -B. Li, Z. -M. Zeren, J. -P. Huang, F. -Q. Huang, S. Che, Z. -M. Zou, P. Xiong, J. Liu, L. -Q. Zhang, Q. Guo, I. Roth, V. S. Makhmutov , et al. (32 additional authors not shown)

Abstract: A promising perspective is presented that humans can provide hourly warning for strong land earthquakes (EQs, Ms6). Two important atmospheric electrostatic signal features are described. A table that lists 9 strong land EQs with shock time, epicenter, magnitude, weather in the region near the epicenter, precursor beginning time, and precursor duration demonstrates that at approximately several hou… ▽ More A promising perspective is presented that humans can provide hourly warning for strong land earthquakes (EQs, Ms6). Two important atmospheric electrostatic signal features are described. A table that lists 9 strong land EQs with shock time, epicenter, magnitude, weather in the region near the epicenter, precursor beginning time, and precursor duration demonstrates that at approximately several hours to one day before a strong land EQ, the weather conditions are fair near the epicenter, and an abnormal negative atmospheric electrostatic signal is very obvious. Moreover, the mechanism is explained. A method by which someone could determine the epicenter and the magnitude of a forthcoming strong EQ is suggested. Finally, the possibility of realizing hourly warning for strong land EQs in the near future is pointed out. △ Less

Submitted 23 June, 2021; originally announced June 2021.

Federated Multi-View Learning for Private Medical Data Integration and Analysis

Authors: Sicong Che, Hao Peng, Lichao Sun, Yong Chen, Lifang He

Abstract: Along with the rapid expansion of information technology and digitalization of health data, there is an increasing concern on maintaining data privacy while garnering the benefits in medical field. Two critical challenges are identified: Firstly, medical data is naturally distributed across multiple local sites, making it difficult to collectively train machine learning models without data leakage… ▽ More Along with the rapid expansion of information technology and digitalization of health data, there is an increasing concern on maintaining data privacy while garnering the benefits in medical field. Two critical challenges are identified: Firstly, medical data is naturally distributed across multiple local sites, making it difficult to collectively train machine learning models without data leakage. Secondly, in medical applications, data are often collected from different sources and views, resulting in heterogeneity and complexity that requires reconciliation. This paper aims to provide a generic Federated Multi-View Learning (FedMV) framework for multi-view data leakage prevention, which is based on different types of local data availability and enables to accommodate two types of problems: Vertical Federated Multi-View Learning (V-FedMV) and Horizontal Federated Multi-View Learning (H-FedMV). We experimented with real-world keyboard data collected from BiAffect study. The results demonstrated that the proposed FedMV approach can make full use of multi-view data in a privacy-preserving way, and both V-FedMV and H-FedMV methods perform better than their single-view and pairwise counterparts. Besides, the proposed model can be easily adapted to deal with multi-view sequential data in a federated environment, which has been modeled and experimentally studied. To the best of our knowledge, this framework is the first to consider both vertical and horizontal diversification in the multi-view setting, as well as their sequential federated learning. △ Less

Submitted 4 May, 2021; originally announced May 2021.

Comments: 22 pages, 6 figures, journal

ACM Class: I.2; J.3; I.5.2

Equilibration and Filtering of Quantum Hall Edge States in Few-Layer Black Phosphorus

Authors: J. Yang, K. Wang, S. Che, Z. J. Tuchfeld, K. Watanabe, T. Taniguchi, D. Shcherbakov, S. Moon, D. Smirnov, R. Chen, M. Bockrath, C. N. Lau

Abstract: We realize p-p'-p junctions in few-layer black phosphorus (BP) devices, and use magneto-transport measurements to study the equilibration and transmission of edge states at the interfaces of regions with different charge densities. We observe both full equilibration, where all edge channels equilibrate and are equally partitioned at the interfaces, and partial equilibration, where only equilibrati… ▽ More We realize p-p'-p junctions in few-layer black phosphorus (BP) devices, and use magneto-transport measurements to study the equilibration and transmission of edge states at the interfaces of regions with different charge densities. We observe both full equilibration, where all edge channels equilibrate and are equally partitioned at the interfaces, and partial equilibration, where only equilibration only takes place among modes of the same spin polarization. Furthermore, the inner p'-region with low-doping level in the junction can function as a filter for highly doped p-regions which demonstrates gate-tunable transmission of edge channels. △ Less

Submitted 1 December, 2020; originally announced December 2020.

Helical Edge States and Quantum Phase Transitions in Tetralayer Graphene

Authors: Shi Che, Yanmeng Shi, Jiawei Yang, Haidong Tian, Ruoyu Chen, Takashi Taniguchi, Kenji Watanabe, Dmitry Smirnov, Chun Ning Lau, Efrat Shimshoni, Ganpathy Murthy, Herbert A. Fertig

Abstract: Helical conductors with spin-momentum locking are promising platforms for Majorana fermions. Here we report observation of two topologically distinct phases supporting helical edge states in charge neutral Bernal-stacked tetralayer graphene in Hall bar and Corbino geometries. As the magnetic field B and out-of-plane displacement field D are varied, we observe a phase diagram consisting of an insul… ▽ More Helical conductors with spin-momentum locking are promising platforms for Majorana fermions. Here we report observation of two topologically distinct phases supporting helical edge states in charge neutral Bernal-stacked tetralayer graphene in Hall bar and Corbino geometries. As the magnetic field B and out-of-plane displacement field D are varied, we observe a phase diagram consisting of an insulating phase and two metallic phases, with 0, 1 and 2 helical edge states, respectively. These phases are accounted for by a theoretical model that relates their conductance to spin-polarization plateaus. Transitions between them arise from a competition among inter-layer hopping, electrostatic and exchange interaction energies. Our work highlights the complex competing symmetries and the rich quantum phases in few-layer graphene. △ Less

Submitted 16 June, 2020; originally announced June 2020.

Comments: Accepted by PRL

Journal ref: Phys. Rev. Lett. 125, 036803 (2020)

Gate Tunable Magnetism and Giant Magnetoresistance in ABC-stacked Few-Layer Graphene

Authors: Yongjin Lee, Shi Che, Jairo Velasco Jr., David Tran, Jacopo Baima, Francesco Mauri, Matteo Calandra, Marc Bockrath, Chun Ning Lau

Abstract: Magnetism is a prototypical phenomenon of quantum collective state, and has found ubiquitous applications in semiconductor technologies such as dynamic random access memory (DRAM). In conventional materials, it typically arises from the strong exchange interaction among the magnetic moments of d- or f-shell electrons. Magnetism, however, can also emerge in perfect lattices from non-magnetic elemen… ▽ More Magnetism is a prototypical phenomenon of quantum collective state, and has found ubiquitous applications in semiconductor technologies such as dynamic random access memory (DRAM). In conventional materials, it typically arises from the strong exchange interaction among the magnetic moments of d- or f-shell electrons. Magnetism, however, can also emerge in perfect lattices from non-magnetic elements. For instance, flat band systems with high density of states (DOS) may develop spontaneous magnetic ordering, as exemplified by the Stoner criterion. Here we report tunable magnetism in rhombohedral-stacked few-layer graphene (r-FLG). At small but finite doping (n~10^11 cm-2), we observe prominent conductance hysteresis and giant magnetoconductance that exceeds 1000% as a function of magnetic fields. Both phenomena are tunable by density and temperature, and disappears for n>10^12 cm-2 or T>5K. These results are confirmed by first principles calculations, which indicate the formation of a half-metallic state in doped r-FLG, in which the magnetization is tunable by electric field. Our combined experimental and theoretical work demonstrate that magnetism and spin polarization, arising from the strong electronic interactions in flat bands, emerge in a system composed entirely of carbon atoms. The electric field tunability of magnetism provides promise for spintronics and low energy device engineering. △ Less

Submitted 11 November, 2019; originally announced November 2019.

Quantum Hall Effect Measurement of Spin-Orbit Coupling Strengths in Ultraclean Bilayer Graphene/WSe2 Heterostructures

Authors: Dongying Wang, Shi Che, Guixin Cao, Rui Lyu, Kenji Watanabe, Takashi Taniguchi, Chun Ning Lau, Marc Bockrath

Abstract: We study proximity-induced spin-orbit coupling (SOC) in bilayer graphene/few-layer WSe2 heterostructure devices. Contact mode atomic force microscopy (AFM) cleaning yields ultra-clean interfaces and high-mobility devices. In a perpendicular magnetic field, we measure the quantum Hall effect to determine the Landau level structure in the presence of out-of-plane Ising and in-plane Rashba SOC. A dis… ▽ More We study proximity-induced spin-orbit coupling (SOC) in bilayer graphene/few-layer WSe2 heterostructure devices. Contact mode atomic force microscopy (AFM) cleaning yields ultra-clean interfaces and high-mobility devices. In a perpendicular magnetic field, we measure the quantum Hall effect to determine the Landau level structure in the presence of out-of-plane Ising and in-plane Rashba SOC. A distinct Landau level crossing pattern emerges when tuning the charge density and displacement field independently with dual gates, originating from a layer-selective SOC proximity effect. Analyzing the Landau level crossings and measured inter-Landau level energy gaps yields the proximity induced SOC energy scale. The Ising SOC is ~ 2.2 meV, 100 times higher than the intrinsic SOC in graphene, while its sign is consistent with theories predicting a dependence of SOC on interlayer twist angle. The Rashba SOC is ~15 meV. Finally, we infer the magnetic field dependence of the inter-Landau level Coulomb interactions. These ultraclean bilayer graphene/WSe2 heterostructures provide a high mobility system with the potential to realize novel topological electronic states and manipulate spins in nanostructures. △ Less

Submitted 24 September, 2019; originally announced September 2019.

Comments: Nano Letters, in press

AWB-GCN: A Graph Convolutional Network Accelerator with Runtime Workload Rebalancing

Authors: Tong Geng, Ang Li, Runbin Shi, Chunshu Wu, Tianqi Wang, Yanfei Li, Pouya Haghi, Antonino Tumeo, Shuai Che, Steve Reinhardt, Martin Herbordt

Abstract: Deep learning systems have been successfully applied to Euclidean data such as images, video, and audio. In many applications, however, information and their relationships are better expressed with graphs. Graph Convolutional Networks (GCNs) appear to be a promising approach to efficiently learn from graph data structures, having shown advantages in many critical applications. As with other deep l… ▽ More Deep learning systems have been successfully applied to Euclidean data such as images, video, and audio. In many applications, however, information and their relationships are better expressed with graphs. Graph Convolutional Networks (GCNs) appear to be a promising approach to efficiently learn from graph data structures, having shown advantages in many critical applications. As with other deep learning modalities, hardware acceleration is critical. The challenge is that real-world graphs are often extremely large and unbalanced; this poses significant performance demands and design challenges. In this paper, we propose Autotuning-Workload-Balancing GCN (AWB-GCN) to accelerate GCN inference. To address the issue of workload imbalance in processing real-world graphs, three hardware-based autotuning techniques are proposed: dynamic distribution smoothing, remote switching, and row remapping. In particular, AWB-GCN continuously monitors the sparse graph pattern, dynamically adjusts the workload distribution among a large number of processing elements (up to 4K PEs), and, after converging, reuses the ideal configuration. Evaluation is performed using an Intel D5005 FPGA with five commonly-used datasets. Results show that 4K-PE AWB-GCN can significantly elevate PE utilization by 7.7x on average and demonstrate considerable performance speedups over CPUs (3255x), GPUs (80.3x), and a prior GCN accelerator (5.1x). △ Less

Submitted 10 September, 2020; v1 submitted 23 August, 2019; originally announced August 2019.

Software-defined Design Space Exploration for an Efficient DNN Accelerator Architecture

Authors: Ye Yu, Yingmin Li, Shuai Che, Niraj K. Jha, Weifeng Zhang

Abstract: Deep neural networks (DNNs) have been shown to outperform conventional machine learning algorithms across a wide range of applications, e.g., image recognition, object detection, robotics, and natural language processing. However, the high computational complexity of DNNs often necessitates extremely fast and efficient hardware. The problem gets worse as the size of neural networks grows exponenti… ▽ More Deep neural networks (DNNs) have been shown to outperform conventional machine learning algorithms across a wide range of applications, e.g., image recognition, object detection, robotics, and natural language processing. However, the high computational complexity of DNNs often necessitates extremely fast and efficient hardware. The problem gets worse as the size of neural networks grows exponentially. As a result, customized hardware accelerators have been developed to accelerate DNN processing without sacrificing model accuracy. However, previous accelerator design studies have not fully considered the characteristics of the target applications, which may lead to sub-optimal architecture designs. On the other hand, new DNN models have been developed for better accuracy, but their compatibility with the underlying hardware accelerator is often overlooked. In this article, we propose an application-driven framework for architectural design space exploration of DNN accelerators. This framework is based on a hardware analytical model of individual DNN operations. It models the accelerator design task as a multi-dimensional optimization problem. We demonstrate that it can be efficaciously used in application-driven accelerator architecture design. Given a target DNN, the framework can generate efficient accelerator design solutions with optimized performance and area. Furthermore, we explore the opportunity to use the framework for accelerator configuration optimization under simultaneous diverse DNN applications. The framework is also capable of improving neural network models to best fit the underlying hardware resources. △ Less

Submitted 16 January, 2020; v1 submitted 18 March, 2019; originally announced March 2019.

Correlated Insulating and Superconducting States in Twisted Bilayer Graphene Below the Magic Angle

Authors: Emilio Codecido, Qiyue Wang, Ryan Koester, Shi Che, Haidong Tian, Rui Lv, Son Tran, Kenji Watanabe, Takashi Taniguchi, Fan Zhang, Marc Bockrath, Chun Ning Lau

Abstract: The emergence of flat bands and correlated behaviors in 'magic angle' twisted bilayer graphene (tBLG) has sparked tremendous interest, though many aspects of the system are under intense debate. Here we report observation of both superconductivity and the Mott-like insulating state in a tBLG device with a twist angle of approximately 0.93, which is smaller than the magic angle by 15%. At an electr… ▽ More The emergence of flat bands and correlated behaviors in 'magic angle' twisted bilayer graphene (tBLG) has sparked tremendous interest, though many aspects of the system are under intense debate. Here we report observation of both superconductivity and the Mott-like insulating state in a tBLG device with a twist angle of approximately 0.93, which is smaller than the magic angle by 15%. At an electron concentration of +/-5 electrons per moire unit cell, we observe a narrow resistance peak with an activation energy gap of approximately 0.1 meV, indicating the existence of an additional correlated insulating state. This is consistent with theory predicting the presence of a high-energy band with an energetically flat dispersion. At a doping of +/-12 electrons per moire unit cell we observe a resistance peak due to the presence of Dirac points in the spectrum. Our results reveal that the magic range of tBLG is in fact larger than what is previously expected, and provide a wealth of new information to help decipher the strongly correlated phenomena observed in tBLG. △ Less

Submitted 13 February, 2019; originally announced February 2019.

Hardware-Guided Symbiotic Training for Compact, Accurate, yet Execution-Efficient LSTM

Authors: Hongxu Yin, Guoyang Chen, Yingmin Li, Shuai Che, Weifeng Zhang, Niraj K. Jha

Abstract: Many long short-term memory (LSTM) applications need fast yet compact models. Neural network compression approaches, such as the grow-and-prune paradigm, have proved to be promising for cutting down network complexity by skipping insignificant weights. However, current compression strategies are mostly hardware-agnostic and network complexity reduction does not always translate into execution effi… ▽ More Many long short-term memory (LSTM) applications need fast yet compact models. Neural network compression approaches, such as the grow-and-prune paradigm, have proved to be promising for cutting down network complexity by skipping insignificant weights. However, current compression strategies are mostly hardware-agnostic and network complexity reduction does not always translate into execution efficiency. In this work, we propose a hardware-guided symbiotic training methodology for compact, accurate, yet execution-efficient inference models. It is based on our observation that hardware may introduce substantial non-monotonic behavior, which we call the latency hysteresis effect, when evaluating network size vs. inference latency. This observation raises question about the mainstream smaller-dimension-is-better compression strategy, which often leads to a sub-optimal model architecture. By leveraging the hardware-impacted hysteresis effect and sparsity, we are able to achieve the symbiosis of model compactness and accuracy with execution efficiency, thus reducing LSTM latency while increasing its accuracy. We have evaluated our algorithms on language modeling and speech recognition applications. Relative to the traditional stacked LSTM architecture obtained for the Penn Treebank dataset, we reduce the number of parameters by 18.0x (30.5x) and measured run-time latency by up to 2.4x (5.2x) on Nvidia GPUs (Intel Xeon CPUs) without any accuracy degradation. For the DeepSpeech2 architecture obtained for the AN4 dataset, we reduce the number of parameters by 7.0x (19.4x), word error rate from 12.9% to 9.9% (10.4%), and measured run-time latency by up to 1.7x (2.4x) on Nvidia GPUs (Intel Xeon CPUs). Thus, our method yields compact, accurate, yet execution-efficient inference models. △ Less

Submitted 30 January, 2019; originally announced January 2019.

Quantum Parity Hall effect in ABA Graphene

Authors: Petr Stepanov, Yafis Barlas, Shi Che, Kevin Myhro, Greyson Voigt, Ziqi Pi, Kenji Watanabe, Takashi Taniguchi, Dmitry Smirnov, Fan Zhang, R. Lake, Allan MacDonald, Chun Ning Lau

Abstract: The celebrated phenomenon of quantum Hall effect has recently been generalized from transport of conserved charges to that of other approximately conserved state variables, including spin and valley, which are characterized by spin- or valley-polarized boundary states with different chiralities. Here, we report a new class of quantum Hall effect in ABA-stacked graphene trilayers (TLG), the quantum… ▽ More The celebrated phenomenon of quantum Hall effect has recently been generalized from transport of conserved charges to that of other approximately conserved state variables, including spin and valley, which are characterized by spin- or valley-polarized boundary states with different chiralities. Here, we report a new class of quantum Hall effect in ABA-stacked graphene trilayers (TLG), the quantum parity Hall (QPH) effect, in which boundary channels are distinguished by even or odd parity under the systems mirror reflection symmetry. At the charge neutrality point and a small perpendicular magnetic field $B_{\perp}$, the longitudinal conductance $σしぐま_{xx}$ is first quantized to $4e^2/h$, establishing the presence of four edge channels. As $B_{\perp}$ increases, $σしぐま_{xx}$ first decreases to $2e^2/h$, indicating spin-polarized counter-propagating edge states, and then to approximately $0$. These behaviors arise from level crossings between even and odd parity bulk Landau levels, driven by exchange interactions with the underlying Fermi sea, which favor an ordinary insulator ground state in the strong $B_{\perp}$ limit, and a spin-polarized state at intermediate fields. The transitions between spin-polarized and unpolarized states can be tuned by varying Zeeman energy. Our findings demonstrate a topological phase that is protected by a gate-controllable symmetry and sensitive to Coulomb interactions. △ Less

Submitted 7 January, 2019; originally announced January 2019.

Integer and Fractional Quantum Hall effect in Ultra-high Quality Few-layer Black Phosphorus Transistors

Authors: Jiawei Yang, Son Tran, Jason Wu, Shi Che, Petr Stepanov, Takashi Taniguchi, Kenji Watanabe, Hongwoo Baek, Dmitry Smirnov, Ruoyu Chen, Chun Ning Lau

Abstract: As a high mobility two-dimensional semiconductor with strong structural and electronic anisotropy, atomically thin black phosphorus (BP) provides a new playground for investigating the quantum Hall (QH) effect, including outstanding questions such as the functional dependence of Landau level (LL) gaps on magnetic field B, and possible anisotropic fractional QH states. Using encapsulating few-layer… ▽ More As a high mobility two-dimensional semiconductor with strong structural and electronic anisotropy, atomically thin black phosphorus (BP) provides a new playground for investigating the quantum Hall (QH) effect, including outstanding questions such as the functional dependence of Landau level (LL) gaps on magnetic field B, and possible anisotropic fractional QH states. Using encapsulating few-layer BP transistors with mobility up to 55,000 cm2/Vs, we extract LL gaps over an exceptionally wide range of B for QH states at filling factors νにゅー=-1 to -4, which are determined to be linear in B, thus resolving a controversy raised by its anisotropy. Furthermore, a fractional QH state at νにゅー~ -4/3 and an additional feature at -0.56+/- 0.1 are observed, underscoring BP as a tunable 2D platform for exploring electron interactions. △ Less

Submitted 6 April, 2018; originally announced April 2018.

Journal ref: Nano Lett. 18, 1, 229-234 (2018)

Twist Angle-Dependent Bands and Valley Inversion in 2D Materials/hBN Heterostructures

Authors: Shi Che, Petr Stepanov, Supeng Ge, Yongjin Lee, Kevin Myhro, Yanmeng Shi, Ruoyu Chen, Ziqi Pi, Cheng Pan, Bin Cheng, Takashi Taniguchi, Kenji Watanabe, Marc Bockrath, Yafis Barlas, Roger Lake, Chun Ning Lau

Abstract: The use of relative twist angle between adjacent atomic layers in a van der Waals heterostructure, has emerged as a new degree of freedom to tune electronic and optoelectronic properties of devices based on 2D materials. Using ABA-stacked trilayer (TLG) graphene as the model system, we show that, contrary to conventional wisdom, the band structures of 2D materials are systematically tunable depend… ▽ More The use of relative twist angle between adjacent atomic layers in a van der Waals heterostructure, has emerged as a new degree of freedom to tune electronic and optoelectronic properties of devices based on 2D materials. Using ABA-stacked trilayer (TLG) graphene as the model system, we show that, contrary to conventional wisdom, the band structures of 2D materials are systematically tunable depending on their relative alignment angle between hexagonal BN (hBN), even at very large twist angles. Moreover, addition or removal of the hBN substrate results in an inversion of the K and K' valley in TLG's lowest Landau level (LL). Our work illustrates the critical role played by substrates in van der Waals heterostructures and opens the door towards band structure modification and valley control via substrate and twist angle engineering. △ Less

Submitted 9 March, 2018; originally announced March 2018.

Journal ref: Phys. Rev. Lett. 125, 246401 (2020)

Large tunable intrinsic gap in rhombohedral-stacked tetralayer graphene

Authors: K. Myhro, S. Che, Y. Shi, Y. Lee, K. Thilahar, K. Bleich, Dmitry Smirnov, C. N. Lau

Abstract: In rhombohedral-stacked few-layer graphene, the very flat energy bands near the charge neutrality point are unstable to electronic interactions, giving rise to states with spontaneous broken symmetries. Using transport measurements on suspended rhombohedral-stacked tetralayer graphene, we observe an insulating ground state with a large interaction-induced gap up to 80 meV. This gapped state can be… ▽ More In rhombohedral-stacked few-layer graphene, the very flat energy bands near the charge neutrality point are unstable to electronic interactions, giving rise to states with spontaneous broken symmetries. Using transport measurements on suspended rhombohedral-stacked tetralayer graphene, we observe an insulating ground state with a large interaction-induced gap up to 80 meV. This gapped state can be enhanced by a perpendicular magnetic field, and suppressed by an interlayer potential, carrier density, or a critical temperature of ~ 40 K. △ Less

Submitted 8 March, 2018; originally announced March 2018.

Tunable Lifshitz Transitions and Multiband Transport in Tetralayer Graphene

Authors: Yanmeng Shi, Shi Che, Kuan Zhou, Supeng Ge, Ziqi Pi, Timothy Espiritu, Takashi Taniguchi, Kenji Watanabe, Yafis Barlas, Roger Lake, Chun Ning Lau

Abstract: As the Fermi level and band structure of two-dimensional materials are readily tunable, they constitute an ideal platform for exploring Lifshitz transition, a change in the topology of a material's Fermi surface. Using tetralayer graphene that host two intersecting massive Dirac bands, we demonstrate multiple Lifshitz transitions and multiband transport, which manifest as non-monotonic dependence… ▽ More As the Fermi level and band structure of two-dimensional materials are readily tunable, they constitute an ideal platform for exploring Lifshitz transition, a change in the topology of a material's Fermi surface. Using tetralayer graphene that host two intersecting massive Dirac bands, we demonstrate multiple Lifshitz transitions and multiband transport, which manifest as non-monotonic dependence of conductivity on charge density n and out-of-plane electric fieldD, anomalous quantum Hall sequences and Landau level crossings that evolve with n, D and B. △ Less

Submitted 20 February, 2018; originally announced February 2018.

Comments: accepted to PRL

Long-Distance Spin Transport Through a Graphene Quantum Hall Antiferromagnet

Authors: Petr Stepanov, Shi Che, Dmitry Shcherbakov, Jiawei Yang, Kevin Thilahar, Greyson Voigt, Marc W. Bockrath, Dmitry Smirnov, Kenji Watanabe, Takashi Taniguchi, Roger K. Lake, Yafis Barlas, Allan H. MacDonald, Chun Ning Lau

Abstract: Antiferromagnetic insulators (AFMI) are robust against stray fields, and their intrinsic dynamics could enable ultrafast magneto-optics and ultrascaled magnetic information processing. Low dissipation, long distance spin transport and electrical manipulation of antiferromagnetic order are much sought-after goals of spintronics research. Here, we report the first experimental evidence of robust lon… ▽ More Antiferromagnetic insulators (AFMI) are robust against stray fields, and their intrinsic dynamics could enable ultrafast magneto-optics and ultrascaled magnetic information processing. Low dissipation, long distance spin transport and electrical manipulation of antiferromagnetic order are much sought-after goals of spintronics research. Here, we report the first experimental evidence of robust long-distance spin transport through an AFMI, in our case the gate-controlled, canted antiferromagnetic (CAF) state that appears at the charge neutrality point of graphene in the presence of an external magnetic field. Utilizing gate-controlled quantum Hall (QH) edge states as spin-dependent injectors and detectors, we observe large, non-local electrical signals across a 5 micron-long, insulating channel only when it is biased into the nu=0 CAF state. Among possible transport mechanisms, spin superfluidity in an antiferromagnetic state gives the most consistent interpretation of the non-local signal's dependence on magnetic field, temperature and filling factors. This work also demonstrates that graphene in the QH regime is a powerful model system for fundamental studies of antiferromagnetic, and in the case of a large in-plane field, ferromagnetic spintronics. △ Less

Submitted 22 January, 2018; originally announced January 2018.

Journal ref: Nature Physics (2018)

Unusual interlayer quantum transport behavior caused by the zeroth Landau level in YbMnBi2

Authors: J. Y. Liu, J. Hu, D. Graf, T. Zou, M. Zhu, Y. Shi, S. Che, S. M. A. Radmanesh, C. N. Lau, L. Spinu, H. B. Cao, X. Ke, Z. Q. Mao

Abstract: Relativistic fermions in topological quantum materials are characterized by linear energy-momentum dispersion near band crossing points. Under magnetic field, relativistic fermions acquire Berry phase of πぱい in cyclotron motion, leading to a zeroth Landau level (LL) at the crossing point. Such field-independent zeroth LL, which distinguishes relativistic fermions from conventional electron systems,… ▽ More Relativistic fermions in topological quantum materials are characterized by linear energy-momentum dispersion near band crossing points. Under magnetic field, relativistic fermions acquire Berry phase of πぱい in cyclotron motion, leading to a zeroth Landau level (LL) at the crossing point. Such field-independent zeroth LL, which distinguishes relativistic fermions from conventional electron systems, is hardly probed in transport measurements since the Fermi energy (EF) is usually not right at the band crossing points in most topological materials. Here we report the observation of exotic quantum transport behavior resulting from the zeroth LL in a multiband topological semimetal YbMnBi2 which possesses linear band crossings both at and away from the Fermi level (FL). We show that the Dirac bands with the crossing points being above or below the FL leads to Shubnikov de-Haas oscillations in the in-plane magnetoresistance, whereas the Dirac bands with the crossing points being at the FL results in unusual angular dependences of the out-of-plane magnetoresistance and in-plane Hall resistivity due to the dependence of the zeroth LL's degeneracy on field orientation. Our results shed light on the transport mechanism of the zeroth LL's relativistic fermions in layered materials. △ Less

Submitted 21 August, 2016; originally announced August 2016.

Journal ref: Nature Communications 8, 646 (2017)

Tunable symmetries of integer and fractional quantum Hall phases in heterostructures with multiple Dirac bands

Authors: Petr Stepanov, Yafis Barlas, Tim Espiritu, Shi Che, Kenji Watanabe, Takashi Taniguchi, Dmitry Smirnov, Chun Ning Lau

Abstract: The co-presence of multiple Dirac bands in few-layer graphene leads to a rich phase diagram in the quantum Hall regime. Using transport measurements, we map the phase diagram of BN-encapsulated ABA-stacked trilayer graphene as a function charge density n, magnetic field B and interlayer displacement field D, and observe transitions among states with different spin, valley, orbital and parity polar… ▽ More The co-presence of multiple Dirac bands in few-layer graphene leads to a rich phase diagram in the quantum Hall regime. Using transport measurements, we map the phase diagram of BN-encapsulated ABA-stacked trilayer graphene as a function charge density n, magnetic field B and interlayer displacement field D, and observe transitions among states with different spin, valley, orbital and parity polarizations. Such rich pattern arises from crossings between Landau levels from different subbands, which reflect the evolving symmetries that are tunable in situ. At D=0, we observe fractional quantum Hall (FQH) states at filling factors 2/3 and -11/3. Unlike those in bilayer graphene, these FQH states are destabilized by a small interlayer potential that hybridizes the different Dirac bands. △ Less

Submitted 19 July, 2016; v1 submitted 13 July, 2016; originally announced July 2016.

Comments: accepted by PRL; corrected author list

Topological nodal-line fermions in ZrSiSe and ZrSiTe

Authors: Jin Hu, Zhijie Tang, Jinyu Liu, Xue Liu, Yanglin Zhu, David Graf, Yanmeng Shi, Shi Che, Chun Ning Lau, Jiang Wei, Zhiqiang Mao

Abstract: The discovery of topological semimetal phase in three-dimensional (3D) systems is a new breakthrough in topological material research. Dirac nodal-line semimetal is one of the three topological semimetal phases discovered so far; it is characterized by linear band crossing along a line/loop, contrasted with the linear band crossing at discrete momentum points in 3D Dirac and Weyl semimetals. The s… ▽ More The discovery of topological semimetal phase in three-dimensional (3D) systems is a new breakthrough in topological material research. Dirac nodal-line semimetal is one of the three topological semimetal phases discovered so far; it is characterized by linear band crossing along a line/loop, contrasted with the linear band crossing at discrete momentum points in 3D Dirac and Weyl semimetals. The study of nodal-line semimetal is still at initial stage; only three material systems have been verified to host nodal line fermions until now, including PbTaSe2, PtSn 4and ZrSiS. In this letter, we report evidence of nodal line fermions in ZrSiSe and ZrSiTe probed in de Haas - van Alphen (dHvA) quantum oscillations. Although ZrSiSe and ZrSiTe share similar layered structure with ZrSiS, our measurements of angular dependences of dHvA oscillations indicate the Fermi surface (FS) enclosing Dirac nodal line is of 2D character in ZiSiTe, in contrast with 3D-like FS in ZrSiSe and ZrSiS. Another important property revealed in our experiment is that the nodal line fermion density in ZrSi(S/Se) (~ 10^20-10^21 cm^-3) is much higher than the Dirac/Weyl fermion density of any known topological materials. In addition, we have demonstrated ZrSiSe and ZrSiTe single crystals can be thinned down to 2D atomic thin layers through microexfoliation, which offers a promising platform to verify the predicted 2D topological insulator in the monolayer materials with ZrSiS-type structure △ Less

Submitted 23 April, 2016; originally announced April 2016.

Journal ref: Phys. Rev. Lett. 117, 016602 (2016)

Energy gaps and layer polarization of integer and fractional quantum Hall states in bilayer graphene

Authors: Yanmeng Shi, Yongjin Lee, Shi Che, Ziqi Pi, Timothy Espiritu, Petr Stepanov, Dmitry Smirnov, Chun Ning Lau, Fan Zhang

Abstract: Owing to the spin, valley, and orbital symmetries, the lowest Landau level (LL) in bilayer graphene exhibits multicomponent quantum Hall ferromagnetism. Using transport spectroscopy, we investigate the energy gaps of integer and fractional quantum Hall states in bilayer graphene with controlled layer polarization. The state at filling factor νにゅー=1 has two distinct phases: a layer polarized state tha… ▽ More Owing to the spin, valley, and orbital symmetries, the lowest Landau level (LL) in bilayer graphene exhibits multicomponent quantum Hall ferromagnetism. Using transport spectroscopy, we investigate the energy gaps of integer and fractional quantum Hall states in bilayer graphene with controlled layer polarization. The state at filling factor νにゅー=1 has two distinct phases: a layer polarized state that has a larger energy gap and is stabilized by high electric field, and a hitherto unobserved interlayer coherent state with a smaller gap that is stabilized by large magnetic field. In contrast, the νにゅー=2/3 quantum Hall state and a feature at νにゅー=1/2 are only resolved at finite electric field and large magnetic field. These results underscore the importance of controlling layer polarization in understanding the competing symmetries in the unusual QH system of BLG. △ Less

Submitted 15 January, 2016; originally announced January 2016.