-
DeepMPR: Enhancing Opportunistic Routing in Wireless Networks through Multi-Agent Deep Reinforcement Learning
Authors:
Saeed Kaviani,
Bo Ryu,
Ejaz Ahmed,
Deokseong Kim,
Jae Kim,
Carrie Spiker,
Blake Harnden
Abstract:
Opportunistic routing relies on the broadcast capability of wireless networks. It brings higher reliability and robustness in highly dynamic and/or severe environments such as mobile or vehicular ad-hoc networks (MANETs/VANETs). To reduce the cost of broadcast, multicast routing schemes use the connected dominating set (CDS) or multi-point relaying (MPR) set to decrease the network overhead and he…
▽ More
Opportunistic routing relies on the broadcast capability of wireless networks. It brings higher reliability and robustness in highly dynamic and/or severe environments such as mobile or vehicular ad-hoc networks (MANETs/VANETs). To reduce the cost of broadcast, multicast routing schemes use the connected dominating set (CDS) or multi-point relaying (MPR) set to decrease the network overhead and hence, their selection algorithms are critical. Common MPR selection algorithms are heuristic, rely on coordination between nodes, need high computational power for large networks, and are difficult to tune for network uncertainties. In this paper, we use multi-agent deep reinforcement learning to design a novel MPR multicast routing technique, DeepMPR, which is outperforming the OLSR MPR selection algorithm while it does not require MPR announcement messages from the neighbors. Our evaluation results demonstrate the performance gains of our trained DeepMPR multicast forwarding policy compared to other popular techniques.
△ Less
Submitted 16 June, 2023;
originally announced June 2023.
-
Open Source-based Over-The-Air 5G New Radio Sidelink Testbed
Authors:
Melissa Elkadi,
Doekseong Kim,
Ejaz Ahmed,
Moein Sadeghi,
Anh Le,
Paul Russell,
Bo Ryu
Abstract:
The focus of this paper is to demonstrate an over-the-air (OTA) 5G new radio (NR) sidelink communication prototype. 5G NR sidelink communications allow NR UEs to transfer data independently without the assistance of a base station (gNB), which enables V2X communications, including platooning, autonomous driving, sensor extension, industrial IoT, public safety communication and much more. Our desig…
▽ More
The focus of this paper is to demonstrate an over-the-air (OTA) 5G new radio (NR) sidelink communication prototype. 5G NR sidelink communications allow NR UEs to transfer data independently without the assistance of a base station (gNB), which enables V2X communications, including platooning, autonomous driving, sensor extension, industrial IoT, public safety communication and much more. Our design leverages the open-source OpenAirInterface5G (OAI) software, which operates on software-defined radios (SDRs) and can be easily extended for mesh networking. The software includes all signal processing components specified by the 3GPP 5G sidelink standards, including Low-Density Parity Check (LDPC) encoding/decoding, polar encoding/decoding, data and control multiplexing, modulation/demodulation, and orthogonal frequency-division multiplexing (OFDM) modulation/demodulation. It can be configured to operate with different bands, bandwidths, and antenna settings. The first milestone in this work was to demonstrate the completed Physical Sidelink Broadcast Channel (PSBCH) development, which conducts synchronization between a Synchronization Reference (SyncRef) UE and a nearby UE. The SyncRef UE broadcasts a sidelink synchronization signal block (S-SSB) periodically, which the nearby UE detects and uses to synchronize its timing and frequency components with the SyncRef UE. Once a connection is established, the next developmental milestone is to transmit real data (text messages) via the Physical Sidelink Shared Channel (PSSCH). Our PHY sidelink framework is tested using both an RF simulator and an OTA testbed with multiple nearby UEs. Beyond the development of synchronization and data transmission/reception in 5G sidelink, we conclude with various performance tests and validation experiments. The results of these metrics show that our simulator is comparable to the OTA testbed.
△ Less
Submitted 6 October, 2023; v1 submitted 15 June, 2023;
originally announced June 2023.
-
DeepADMR: A Deep Learning based Anomaly Detection for MANET Routing
Authors:
Alex Yahja,
Saeed Kaviani,
Bo Ryu,
Jae H. Kim,
Kevin A. Larson
Abstract:
We developed DeepADMR, a novel neural anomaly detector for the deep reinforcement learning (DRL)-based DeepCQ+ MANET routing policy. The performance of DRL-based algorithms such as DeepCQ+ is only verified within the trained and tested environments, hence their deployment in the tactical domain induces high risks. DeepADMR monitors unexpected behavior of the DeepCQ+ policy based on the temporal di…
▽ More
We developed DeepADMR, a novel neural anomaly detector for the deep reinforcement learning (DRL)-based DeepCQ+ MANET routing policy. The performance of DRL-based algorithms such as DeepCQ+ is only verified within the trained and tested environments, hence their deployment in the tactical domain induces high risks. DeepADMR monitors unexpected behavior of the DeepCQ+ policy based on the temporal difference errors (TD-errors) in real-time and detects anomaly scenarios with empirical and non-parametric cumulative-sum statistics. The DeepCQ+ design via multi-agent weight-sharing proximal policy optimization (PPO) is slightly modified to enable the real-time estimation of the TD-errors. We report the DeepADMR performance in the presence of channel disruptions, high mobility levels, and network sizes beyond the training environments, which shows its effectiveness.
△ Less
Submitted 24 January, 2023;
originally announced February 2023.
-
Best Thermoelectric Efficiency of Ever-Explored Materials
Authors:
Byungki Ryu,
Jaywan Chung,
Masaya Kumagai,
Tomoya Mato,
Yuki Ando,
Sakiko Gunji,
Atsumi Tanaka,
Dewi Yana,
Masayuki Fujimoto,
Yoji Imai,
Yukari Katsura,
SuDong Park
Abstract:
A thermoelectric device is a heat engine that directly converts heat into electricity. Many materials with a high figure of merit ZT have been discovered in anticipation of a high thermoelectric efficiency. However, there has been a lack of investigations on efficiency-based material evaluation, and little is known about the achievable limit of thermoelectric efficiency. Here, we report the highes…
▽ More
A thermoelectric device is a heat engine that directly converts heat into electricity. Many materials with a high figure of merit ZT have been discovered in anticipation of a high thermoelectric efficiency. However, there has been a lack of investigations on efficiency-based material evaluation, and little is known about the achievable limit of thermoelectric efficiency. Here, we report the highest thermoelectric efficiency using 12,645 published materials. The 97,841,810 thermoelectric efficiencies are calculated using 808,610 device configurations under various heat-source temperatures (T_h) when the cold-side temperature is 300 K, solving one-dimensional thermoelectric integral equations with temperature-dependent thermoelectric properties. For infinite-cascade devices, a thermoelectric efficiency larger than 33% (~1/3) is achievable when T_h exceeds 1400 K. For single-stage devices, the best efficiency of 17.1% (~1/6) is possible when T_h is 860 K. Leg segmentation can overcome this limit, delivering a very high efficiency of 24% (~1/4) when T_h is 1100 K.
△ Less
Submitted 14 March, 2023; v1 submitted 17 October, 2022;
originally announced October 2022.
-
Deep Reinforcement Learning for System-on-Chip: Myths and Realities
Authors:
Tegg Taekyong Sung,
Bo Ryu
Abstract:
Neural schedulers based on deep reinforcement learning (DRL) have shown considerable potential for solving real-world resource allocation problems, as they have demonstrated significant performance gain in the domain of cluster computing. In this paper, we investigate the feasibility of neural schedulers for the domain of System-on-Chip (SoC) resource allocation through extensive experiments and c…
▽ More
Neural schedulers based on deep reinforcement learning (DRL) have shown considerable potential for solving real-world resource allocation problems, as they have demonstrated significant performance gain in the domain of cluster computing. In this paper, we investigate the feasibility of neural schedulers for the domain of System-on-Chip (SoC) resource allocation through extensive experiments and comparison with non-neural, heuristic schedulers. The key finding is three-fold. First, neural schedulers designed for cluster computing domain do not work well for SoC due to i) heterogeneity of SoC computing resources and ii) variable action set caused by randomness in incoming jobs. Second, our novel neural scheduler technique, Eclectic Interaction Matching (EIM), overcomes the above challenges, thus significantly improving the existing neural schedulers. Specifically, we rationalize the underlying reasons behind the performance gain by the EIM-based neural scheduler. Third, we discover that the ratio of the average processing elements (PE) switching delay and the average PE computation time significantly impacts the performance of neural SoC schedulers even with EIM. Consequently, future neural SoC scheduler design must consider this metric as well as its implementation overhead for practical utility.
△ Less
Submitted 29 July, 2022;
originally announced July 2022.
-
DeepCQ+: Robust and Scalable Routing with Multi-Agent Deep Reinforcement Learning for Highly Dynamic Networks
Authors:
Saeed Kaviani,
Bo Ryu,
Ejaz Ahmed,
Kevin Larson,
Anh Le,
Alex Yahja,
Jae H. Kim
Abstract:
Highly dynamic mobile ad-hoc networks (MANETs) remain as one of the most challenging environments to develop and deploy robust, efficient, and scalable routing protocols. In this paper, we present DeepCQ+ routing protocol which, in a novel manner integrates emerging multi-agent deep reinforcement learning (MADRL) techniques into existing Q-learning-based routing protocols and their variants and ac…
▽ More
Highly dynamic mobile ad-hoc networks (MANETs) remain as one of the most challenging environments to develop and deploy robust, efficient, and scalable routing protocols. In this paper, we present DeepCQ+ routing protocol which, in a novel manner integrates emerging multi-agent deep reinforcement learning (MADRL) techniques into existing Q-learning-based routing protocols and their variants and achieves persistently higher performance across a wide range of topology and mobility configurations. While kee** the overall protocol structure of the Q-learning-based routing protocols, DeepCQ+ replaces statically configured parameterized thresholds and hand-written rules with carefully designed MADRL agents such that no configuration of such parameters is required a priori. Extensive simulation shows that DeepCQ+ yields significantly increased end-to-end throughput with lower overhead and no apparent degradation of end-to-end delays (hop counts) compared to its Q-learning based counterparts. Qualitatively, and perhaps more significantly, DeepCQ+ maintains remarkably similar performance gains under many scenarios that it was not trained for in terms of network sizes, mobility conditions, and traffic dynamics. To the best of our knowledge, this is the first successful application of the MADRL framework for the MANET routing problem that demonstrates a high degree of scalability and robustness even under environments that are outside the trained range of scenarios. This implies that our MARL-based DeepCQ+ design solution significantly improves the performance of Q-learning based CQ+ baseline approach for comparison and increases its practicality and explainability because the real-world MANET environment will likely vary outside the trained range of MANET scenarios. Additional techniques to further increase the gains in performance and scalability are discussed.
△ Less
Submitted 29 November, 2021;
originally announced November 2021.
-
Identifying Two Newly Discovered Black Hole X-ray Transients in the Galactic Center
Authors:
Byeong Yeon Ryu
Abstract:
We analyze the two unclassified newly outbursting X-ray binary transients in the galactic center, Swift J174540.7-290015 and Swift J174540.2-290037, in an attempt to identify the nature of these sources. We engage in a thorough spectral analysis for the two systems, by examining the fits of various astrophysical models, in addition to assessing their light curves and power spectra. Our results sho…
▽ More
We analyze the two unclassified newly outbursting X-ray binary transients in the galactic center, Swift J174540.7-290015 and Swift J174540.2-290037, in an attempt to identify the nature of these sources. We engage in a thorough spectral analysis for the two systems, by examining the fits of various astrophysical models, in addition to assessing their light curves and power spectra. Our results showed that Swift 15 has a low temperature blackbody spectrum with a hard power-law tail, and Swift 37 has a broad iron line and a small equivalent width; both transients exhibit classic spectral signatures of black hole systems. Additionally, our observations for both sources yielded the absence of type 1 X-ray bursting and the possible presence of low-frequency quasi-periodic oscillations. Our finding that these two sources appear to be black hole X-ray systems is a groundbreaking discovery, as it adds on to an extremely small population of well-established black hole candidates in the entire galaxy, in addition to giving the first strong indication for a clustering of black holes within the central few parsecs of the galactic center.
△ Less
Submitted 2 November, 2021;
originally announced November 2021.
-
Unique Temperature Distribution and Explicit Efficiency Formula for One-Dimensional Thermoelectric Generators under Constant Seebeck Coefficients
Authors:
Jaywan Chung,
Byungki Ryu,
Hyowon Seo
Abstract:
A thermoelectric generator converts a temperature difference into electrical energy. Its energy conversion efficiency is determined by the steady-state temperature distribution inside the generator. By assuming the thermoelectric material in the generator has a temperature-independent Seebeck coefficient and the generator is one-dimensional, we show that the second-order integro-differential equat…
▽ More
A thermoelectric generator converts a temperature difference into electrical energy. Its energy conversion efficiency is determined by the steady-state temperature distribution inside the generator. By assuming the thermoelectric material in the generator has a temperature-independent Seebeck coefficient and the generator is one-dimensional, we show that the second-order integro-differential equation describing the inside temperature distribution has a unique solution for any given ratio of external load resistance to the internal resistance. Hence the efficiency is well defined. Furthermore, we show the efficiency has an explicit formula in terms of the temperature-dependent thermal conductivity and electrical resistivity of the thermoelectric material. On the other hand, if we impose an external load resistance value, not the ratio, then the integro-differential equation can have multiple solutions.
△ Less
Submitted 21 June, 2021;
originally announced June 2021.
-
SoCRATES: System-on-Chip Resource Adaptive Scheduling using Deep Reinforcement Learning
Authors:
Tegg Taekyong Sung,
Bo Ryu
Abstract:
Deep Reinforcement Learning (DRL) is being increasingly applied to the problem of resource allocation for emerging System-on-Chip (SoC) applications, and has shown remarkable promises. In this paper, we introduce SoCRATES (SoC Resource AdapTivE Scheduler), an extremely efficient DRL-based SoC scheduler which maps a wide range of hierarchical jobs to heterogeneous resources within SoC using the Ecl…
▽ More
Deep Reinforcement Learning (DRL) is being increasingly applied to the problem of resource allocation for emerging System-on-Chip (SoC) applications, and has shown remarkable promises. In this paper, we introduce SoCRATES (SoC Resource AdapTivE Scheduler), an extremely efficient DRL-based SoC scheduler which maps a wide range of hierarchical jobs to heterogeneous resources within SoC using the Eclectic Interaction Matching (EIM) technique. It is noted that the majority of SoC resource management approaches have been targeting makespan minimization with fixed number of jobs in the system. In contrast, SoCRATES aims at minimizing average latency in a steady-state condition while assigning tasks in the ready queue to heterogeneous resources (processing elements). We first show that the latency-minimization-driven SoC applications operate high-frequency job workload and distributed/parallel job execution. We then demonstrate SoCRATES successfully addresses the challenge of concurrent observations caused by the task dependency inherent in the latency minimization objective. Extensive tests show that SoCRATES outperforms other existing neural and non-neural schedulers with as high as 38% gain in latency reduction under a variety of job types and incoming rates. The resulting model is also compact in size and has very favorable energy consumption behaviors, making it highly practical for deployment in future SoC systems with built-in neural accelerator.
△ Less
Submitted 11 October, 2021; v1 submitted 28 April, 2021;
originally announced April 2021.
-
A Scalable and Reproducible System-on-Chip Simulation for Reinforcement Learning
Authors:
Tegg Taekyong Sung,
Bo Ryu
Abstract:
Deep Reinforcement Learning (DRL) underlies in a simulated environment and optimizes objective goals. By extending the conventional interaction scheme, this paper proffers gym-ds3, a scalable and reproducible open environment tailored for a high-fidelity Domain-Specific System-on-Chip (DSSoC) application. The simulation corroborates to schedule hierarchical jobs onto heterogeneous System-on-Chip (…
▽ More
Deep Reinforcement Learning (DRL) underlies in a simulated environment and optimizes objective goals. By extending the conventional interaction scheme, this paper proffers gym-ds3, a scalable and reproducible open environment tailored for a high-fidelity Domain-Specific System-on-Chip (DSSoC) application. The simulation corroborates to schedule hierarchical jobs onto heterogeneous System-on-Chip (SoC) processors and bridges the system to reinforcement learning research. We systematically analyze the representative SoC simulator and discuss the primary challenging aspects that the system (1) continuously generates indefinite jobs at a rapid injection rate, (2) optimizes complex objectives, and (3) operates in steady-state scheduling. We provide exemplary snippets and experimentally demonstrate the run-time performances on different schedulers that successfully mimic results achieved from the standard DS3 framework and real-world embedded systems.
△ Less
Submitted 27 April, 2021;
originally announced April 2021.
-
Robust and Scalable Routing with Multi-Agent Deep Reinforcement Learning for MANETs
Authors:
Saeed Kaviani,
Bo Ryu,
Ejaz Ahmed,
Kevin A. Larson,
Anh Le,
Alex Yahja,
Jae H. Kim
Abstract:
Highly dynamic mobile ad-hoc networks (MANETs) are continuing to serve as one of the most challenging environments to develop and deploy robust, efficient, and scalable routing protocols. In this paper, we present DeepCQ+ routing which, in a novel manner, integrates emerging multi-agent deep reinforcement learning (MADRL) techniques into existing Q-learning-based routing protocols and their varian…
▽ More
Highly dynamic mobile ad-hoc networks (MANETs) are continuing to serve as one of the most challenging environments to develop and deploy robust, efficient, and scalable routing protocols. In this paper, we present DeepCQ+ routing which, in a novel manner, integrates emerging multi-agent deep reinforcement learning (MADRL) techniques into existing Q-learning-based routing protocols and their variants, and achieves persistently higher performance across a wide range of MANET configurations while training only on a limited range of network parameters and conditions. Quantitatively, DeepCQ+ shows consistently higher end-to-end throughput with lower overhead compared to its Q-learning-based counterparts with the overall gain of 10-15% in its efficiency. Qualitatively and more significantly, DeepCQ+ maintains remarkably similar performance gains under many scenarios that it was not trained for in terms of network sizes, mobility conditions, and traffic dynamics. To the best of our knowledge, this is the first successful demonstration of MADRL for the MANET routing problem that achieves and maintains a high degree of scalability and robustness even in the environments that are outside the trained range of scenarios. This implies that the proposed hybrid design approach of DeepCQ+ that combines MADRL and Q-learning significantly increases its practicality and explainability because the real-world MANET environment will likely vary outside the trained range of MANET scenarios.
△ Less
Submitted 28 March, 2021; v1 submitted 8 January, 2021;
originally announced January 2021.
-
Off-centered Pb interstitials in PbTe
Authors:
Sung** Park,
Jongho Park,
Byungki Ryu,
SuDong Park
Abstract:
In this work, we calculate the defect properties of low-symmetry Pb interstitials in PbTe using first-principles density-functional theory calculations. We break the symmetry imposed on on-centered interstitial defects and show that the lowest ground state of Pb interstitial defects is off-centered along the [111] directions. Due to the four multi-stable structures with low defect formation energi…
▽ More
In this work, we calculate the defect properties of low-symmetry Pb interstitials in PbTe using first-principles density-functional theory calculations. We break the symmetry imposed on on-centered interstitial defects and show that the lowest ground state of Pb interstitial defects is off-centered along the [111] directions. Due to the four multi-stable structures with low defect formation energies, the defect density of Pb interstitials is expected to be ~5.6 times larger than previous predictions when PbTe is synthesized at 900 K. In contrast to the on-centered Pbinterstitial, the off-centered Pb interstitials in PbTe can exhibit long-range lattice relaxation toward [111] direction beyond distance of 1 nm, indicating the potential formation of weak local dipoles. This result provides an alternative explanation for the emphanitic anharmonicity of PbTe.
△ Less
Submitted 25 August, 2020;
originally announced August 2020.
-
DeepSoCS: A Neural Scheduler for Heterogeneous System-on-Chip (SoC) Resource Scheduling
Authors:
Tegg Taekyong Sung,
Jeongsoo Ha,
Jeewoo Kim,
Alex Yahja,
Chae-Bong Sohn,
Bo Ryu
Abstract:
In this paper, we~present a novel scheduling solution for a class of System-on-Chip (SoC) systems where heterogeneous chip resources (DSP, FPGA, GPU, etc.) must be efficiently scheduled for continuously arriving hierarchical jobs with their tasks represented by a directed acyclic graph. Traditionally, heuristic algorithms have been widely used for many resource scheduling domains, and Heterogeneou…
▽ More
In this paper, we~present a novel scheduling solution for a class of System-on-Chip (SoC) systems where heterogeneous chip resources (DSP, FPGA, GPU, etc.) must be efficiently scheduled for continuously arriving hierarchical jobs with their tasks represented by a directed acyclic graph. Traditionally, heuristic algorithms have been widely used for many resource scheduling domains, and Heterogeneous Earliest Finish Time (HEFT) has been a dominating state-of-the-art technique across a broad range of heterogeneous resource scheduling domains over many years. Despite their long-standing popularity, HEFT-like algorithms are known to be vulnerable to a small amount of noise added to the environment. Our Deep Reinforcement Learning (DRL)-based SoC Scheduler (DeepSoCS), capable of learning the "best" task ordering under dynamic environment changes, overcomes the brittleness of rule-based schedulers such as HEFT with significantly higher performance across different types of jobs. We~describe a DeepSoCS design process using a real-time heterogeneous SoC scheduling emulator, discuss major challenges, and present two novel neural network design features that lead to outperforming HEFT: (i) hierarchical job- and task-graph embedding; and (ii) efficient use of real-time task information in the state space. Furthermore, we~introduce effective techniques to address two fundamental challenges present in our environment: delayed consequences and joint actions. Through an extensive simulation study, we~show that our DeepSoCS exhibits the significantly higher performance of job execution time than that of HEFT with a higher level of robustness under realistic noise conditions. We~conclude with a discussion of the potential improvements for our DeepSoCS neural scheduler.
△ Less
Submitted 5 June, 2020; v1 submitted 15 May, 2020;
originally announced May 2020.
-
Native point defects and low $p$-do** efficiency in $Mg_2 (Si,Sn)$ solid solutions: A hybrid-density functional study
Authors:
Byungki Ryu,
Eun-Ae Choi,
Sung** Park,
Jaywan Chung,
Johannes de Boor,
Pawel Ziolkowski,
Eckhard Müller,
SuDong Park
Abstract:
We perform hybrid-density functional calculations to investigate the charged defect formation energy of native point defects in $Mg_2 Si$, $Mg_2 Sn$, and their solid solutions. The band gap correction by hybrid-density functional is found to be critical to determine the charged defect density in these materials. For $Mg_2 Si$, $Mg$ interstitials are dominant and provide unintentional $n$-type cond…
▽ More
We perform hybrid-density functional calculations to investigate the charged defect formation energy of native point defects in $Mg_2 Si$, $Mg_2 Sn$, and their solid solutions. The band gap correction by hybrid-density functional is found to be critical to determine the charged defect density in these materials. For $Mg_2 Si$, $Mg$ interstitials are dominant and provide unintentional $n$-type conductivity. Additionally, as the $Mg$ vacancies can dominate in $Mg$-poor $Mg_2 Sn$, $p$-type conductivity is possible for $Mg_2 Sn$. However, the existence of low formation energy defects such as $Mg_{Sn}^{1+}$ and $I_{Mg}^{2+}$ in $Mg_2 Sn$ and their diffusion can cause severe charge compensation of hole carriers resulting in low $p$-type do** efficiency and thermal degradation. Our results indicate that, in addition to the extrinsic do** strategy, alloying of $Mg_2 Si$ with $Mg_2 Sn$ under $Mg$-poor conditions would be necessary to enhance the $p$-type conductivity with less charge compensation.
△ Less
Submitted 14 September, 2020; v1 submitted 19 March, 2020;
originally announced March 2020.
-
Counterintuitive example on relation between ZT and thermoelectric efficiency
Authors:
Byungki Ryu,
Jaywan Chung,
Eun-Ae Choi,
Pawel Ziolkowski,
Eckhard Müller,
SuDong Park
Abstract:
The thermoelectric figure of merit ZT, which is defined using electrical conductivity, Seebeck coefficient, thermal conductivity, and absolute temperature T, has been widely used as a simple estimator of the conversion efficiency of a thermoelectric heat engine. When material properties are constant or slowly varying with T, a higher ZT ensures a higher maximum conversion efficiency of thermoelect…
▽ More
The thermoelectric figure of merit ZT, which is defined using electrical conductivity, Seebeck coefficient, thermal conductivity, and absolute temperature T, has been widely used as a simple estimator of the conversion efficiency of a thermoelectric heat engine. When material properties are constant or slowly varying with T, a higher ZT ensures a higher maximum conversion efficiency of thermoelectric materials. However, as material properties can vary strongly with T, efficiency predictions based on ZT can be inaccurate, especially for wide-temperature applications. Moreover, although ZT values continue to increase, there has been no investigation of the relationship between ZT and the efficiency in the higher ZT regime. In this paper, we report a counterintuitive situation by comparing two materials: although one material has a higher ZT value over the whole operational temperature range, its maximum conversion efficiency is smaller than that of the other. This indicates that, for material comparisons, the evaluation of exact efficiencies as opposed to a simple comparison of the ZTs is necessary in certain cases.
△ Less
Submitted 29 January, 2020;
originally announced January 2020.
-
The Demise of Single-Authored Publications in Computer Science: A Citation Network Analysis
Authors:
Brian K. Ryu
Abstract:
In this study, I analyze the DBLP bibliographic database to study role of single author publications in the computer science literature between 1940 and 2019. I examine the demographics and reception by computing the population fraction, citation statistics, and PageRank scores of single author publications over the years. Both the population fraction and reception have been continuously declining…
▽ More
In this study, I analyze the DBLP bibliographic database to study role of single author publications in the computer science literature between 1940 and 2019. I examine the demographics and reception by computing the population fraction, citation statistics, and PageRank scores of single author publications over the years. Both the population fraction and reception have been continuously declining since the 1940s. The overall decaying trend of single author publications is qualitatively consistent with those observed in other scientific disciplines, though the diminution is taking place several decades later than those in the natural sciences. Additionally, I analyze the scope and volume of single author publications, using page length and reference count as first-order approximations of the scope of publications. Although both metrics on average show positive correlations with citation count, single author papers show no significant difference in page or reference counts compared to the rest of the publications, suggesting that there exist other factors that impact the citations of single author publications.
△ Less
Submitted 2 January, 2020;
originally announced January 2020.
-
General Efficiency Theory of Thermoelectric Conversion
Authors:
Byungki Ryu,
Jaywan Chung,
SuDong Park
Abstract:
In this Letter, we show thermoelectric conversion efficiency is \emph{exactly} determined by \emph{three} independent material parameters $Z_{\rm gen}$, $τ$, and $β$. Each parameter is a figure of merit hence improving $τ$ or $β$ is an additional way to increase the efficiency. The $Z_{\rm gen}$ generalizes the traditional figure of merit \emph{zT}. Two degrees of freedom $τ$ and $β$ reflect the t…
▽ More
In this Letter, we show thermoelectric conversion efficiency is \emph{exactly} determined by \emph{three} independent material parameters $Z_{\rm gen}$, $τ$, and $β$. Each parameter is a figure of merit hence improving $τ$ or $β$ is an additional way to increase the efficiency. The $Z_{\rm gen}$ generalizes the traditional figure of merit \emph{zT}. Two degrees of freedom $τ$ and $β$ reflect the temperature gradients of the material properties and are crucial to evaluate the heat current altered by non-zero Thomson heat and asymmetric Joule heat escape. Physical insights on high $τ$ or $β$ materials explain why the single parameter approaches can be inaccurate for efficiency prediction.
△ Less
Submitted 23 October, 2019;
originally announced October 2019.
-
Neural Heterogeneous Scheduler
Authors:
Tegg Taekyong Sung,
Valliappa Chockalingam,
Alex Yahja,
Bo Ryu
Abstract:
Access to parallel and distributed computation has enabled researchers and developers to improve algorithms and performance in many applications. Recent research has focused on next generation special purpose systems with multiple kinds of coprocessors, known as heterogeneous system-on-chips (SoC). In this paper, we introduce a method to intelligently schedule--and learn to schedule--a stream of t…
▽ More
Access to parallel and distributed computation has enabled researchers and developers to improve algorithms and performance in many applications. Recent research has focused on next generation special purpose systems with multiple kinds of coprocessors, known as heterogeneous system-on-chips (SoC). In this paper, we introduce a method to intelligently schedule--and learn to schedule--a stream of tasks to available processing elements in such a system. We use deep reinforcement learning enabling complex sequential decision making and empirically show that our reinforcement learning system provides for a viable, better alternative to conventional scheduling heuristics with respect to minimizing execution time.
△ Less
Submitted 9 June, 2019;
originally announced June 2019.
-
Hybrid-functional and quasi-particle calculations of band structures of Mg2Si, Mg2Ge, and Mg2Sn
Authors:
Byungki Ryu,
Sung** Park,
Eun-Ae Choi,
Johannes de Boor,
Pawel Ziolkowski,
Jaywan Chung,
SuDong Park
Abstract:
We perform hybrid functional and quasi-particle band structure calculations with spin-orbit interaction to investigate the band structures of Mg2Si, Mg2Ge, and Mg2Sn. For all Mg2X materials, where X = Si, Ge, and Sn, the characteristics of band edge states, i.e., band and valley degeneracies, and orbital characters, are found to be conserved, independent of the computational schemes such as densit…
▽ More
We perform hybrid functional and quasi-particle band structure calculations with spin-orbit interaction to investigate the band structures of Mg2Si, Mg2Ge, and Mg2Sn. For all Mg2X materials, where X = Si, Ge, and Sn, the characteristics of band edge states, i.e., band and valley degeneracies, and orbital characters, are found to be conserved, independent of the computational schemes such as density functional generalized gradient approximation, hybrid functionals, or quasi-particle calculations. However, the magnitude of the calculated band gap varies significantly with the computational schemes. Within density-functional calculations, the one-particle band gaps of Mg2Si, Mg2Ge, and Mg2Sn are 0.191, 0.090, and -0.346 eV, respectively, and thus severely underestimated compared to the experimental gaps, due to the band gap error in the density functional theory and the significant relativistic effect on the low-energy band structures. By employing hybrid-functional calculations with a 35% fraction of the exact Hartree-Fock exchange energy (HSE-35%), we overcame the negative band gap issue in Mg2Sn. Finally, in quasi-particle calculations on top of the HSE-35% Hamiltonians, we obtained band gaps of 0.835, 0.759, and 0.244 eV for Mg2Si, Mg2Ge, and Mg2Sn, respectively, consistent with the experimental band gaps of 0.77, 0.74, and 0.36 eV, respectively.
△ Less
Submitted 15 May, 2019;
originally announced May 2019.
-
Investigation of effective thermoelectric properties of composite with interfacial resistance using micromechanics-based homogenisation
Authors:
Jiyoung Jung,
Sangryun Lee,
Byungki Ryu,
Seunghwa Ryu
Abstract:
We obtained the analytical expression for the effective thermoelectric properties and dimensionless figure of merit of a composite with interfacial electrical and thermal resistances using a micromechanics-based homogenisation. For the first time, we derived the Eshelby tensor for a spherical inclusion as a function of the interfacial resistances and obtained the solutions of the effective Seebeck…
▽ More
We obtained the analytical expression for the effective thermoelectric properties and dimensionless figure of merit of a composite with interfacial electrical and thermal resistances using a micromechanics-based homogenisation. For the first time, we derived the Eshelby tensor for a spherical inclusion as a function of the interfacial resistances and obtained the solutions of the effective Seebeck coefficient and the electrical and thermal conductivities of a composite, which were validated against finite-element analysis (FEA). Our analytical predictions well match the effective properties obtained from FEA with an inclusion volume fraction up to 15%. Because the effective properties were derived with the assumption of a small temperature difference, we discuss a heuristic method for obtaining the effective properties in the case where a thermoelectric composite is subjected to a large temperature difference.
△ Less
Submitted 8 December, 2018; v1 submitted 27 November, 2018;
originally announced November 2018.
-
Thermoelectric degrees of freedom determining thermoelectric efficiency
Authors:
Byungki Ryu,
Jaywan Chung,
SuDong Park
Abstract:
Thermal energy can be directly converted to electrical energy as a result of thermoelectric effects. Because this conversion realises clean energy technology, such as waste heat recovery and energy harvesting, substantial efforts have been made to search for thermoelectric materials. Under the belief that the material figure of merit $zT$ represents the energy conversion efficiencies of thermoelec…
▽ More
Thermal energy can be directly converted to electrical energy as a result of thermoelectric effects. Because this conversion realises clean energy technology, such as waste heat recovery and energy harvesting, substantial efforts have been made to search for thermoelectric materials. Under the belief that the material figure of merit $zT$ represents the energy conversion efficiencies of thermoelectric devices, various high peak-$zT$ materials have been explored for half a century. However, thermoelectric properties vary greatly with temperature $T$, so the single value $zT$ does not represent device efficiency accurately. Here we show that the efficiency of thermoelectric conversion is completely determined by \emph{three} parameters $Z_{\mathrm{gen}}$, $τ$, and $β$, which we call the \emph{thermoelectric degrees of freedom}. The $Z_{\mathrm{gen}}$, which is an average of material properties, is a generalisation of the traditional figure of merit. The $τ$ and $β$, which reflect the gradients of the material properties, are proportional to escaped heat caused by the Thomson effect and asymmetric Joule heat, respectively. Our finding proposes new directions for achieving high thermoelectric efficiency; increasing one of the thermoelectric degrees of freedom results in higher efficiency. For example, thermoelectric efficiency can be enhanced up to 176\% by tuning the thermoelectric degrees of freedom in segmented legs, compared to the best efficiency of single-material legs.
△ Less
Submitted 29 March, 2023; v1 submitted 25 October, 2018;
originally announced October 2018.
-
A micromechanics-based analytical solution for the effective thermal conductivity of composites with orthotropic matrices and interfacial thermal resistance
Authors:
Sangryun Lee,
**yeop Lee,
Byungki Ryu,
Seunghwa Ryu
Abstract:
We obtained an analytical solution for the effective thermal conductivity of composites composed of orthotropic matrices and spherical inhomogeneities with interfacial thermal resistance using a micromechanics-based homogenization. We derived the closed form of a modified Eshelby tensor as a function of the interfacial thermal resistance. We then predicted the heat flux of a single inhomogeneity i…
▽ More
We obtained an analytical solution for the effective thermal conductivity of composites composed of orthotropic matrices and spherical inhomogeneities with interfacial thermal resistance using a micromechanics-based homogenization. We derived the closed form of a modified Eshelby tensor as a function of the interfacial thermal resistance. We then predicted the heat flux of a single inhomogeneity in the infinite media based on the modified Eshelby tensor, which was validated against the numerical results obtained from the finite element analysis. Based on the modified Eshelby tensor and the localization tensor accounting for the interfacial resistance, we derived an analytical expression for the effective thermal conductivity tensor for the composites by a mean-field approach called the Mori-Tanaka method. Our analytical prediction matched very well with the effective thermal conductivity obtained from finite element analysis with up to 10% inhomogeneity volume fraction.
△ Less
Submitted 7 May, 2018; v1 submitted 22 December, 2017;
originally announced December 2017.
-
Highrisk Prediction from Electronic Medical Records via Deep Attention Networks
Authors:
You ** Kim,
Yun-Geun Lee,
Jeong Whun Kim,
** Joo Park,
Borim Ryu,
Jung-Woo Ha
Abstract:
Predicting highrisk vascular diseases is a significant issue in the medical domain. Most predicting methods predict the prognosis of patients from pathological and radiological measurements, which are expensive and require much time to be analyzed. Here we propose deep attention models that predict the onset of the high risky vascular disease from symbolic medical histories sequence of hypertensio…
▽ More
Predicting highrisk vascular diseases is a significant issue in the medical domain. Most predicting methods predict the prognosis of patients from pathological and radiological measurements, which are expensive and require much time to be analyzed. Here we propose deep attention models that predict the onset of the high risky vascular disease from symbolic medical histories sequence of hypertension patients such as ICD-10 and pharmacy codes only, Medical History-based Prediction using Attention Network (MeHPAN). We demonstrate two types of attention models based on 1) bidirectional gated recurrent unit (R-MeHPAN) and 2) 1D convolutional multilayer model (C-MeHPAN). Two MeHPAN models are evaluated on approximately 50,000 hypertension patients with respect to precision, recall, f1-measure and area under the curve (AUC). Experimental results show that our MeHPAN methods outperform standard classification models. Comparing two MeHPANs, R-MeHPAN provides more better discriminative capability with respect to all metrics while C-MeHPAN presents much shorter training time with competitive accuracy.
△ Less
Submitted 30 November, 2017;
originally announced December 2017.
-
Work Function of Bismuth Telluride: First-Principles Approach
Authors:
Byungki Ryu
Abstract:
First-principles approach is demonstrated to calculate the work function of Bi2Te3. The reference potential and the vacuum energy levels are extracted from the Bi2Te3 (0001) surface structure using the reference potential method based on the density functional theory. The one-shot GOWO many-body perturbation theory is used to place the bulk band edge energies with respect to the reference level an…
▽ More
First-principles approach is demonstrated to calculate the work function of Bi2Te3. The reference potential and the vacuum energy levels are extracted from the Bi2Te3 (0001) surface structure using the reference potential method based on the density functional theory. The one-shot GOWO many-body perturbation theory is used to place the bulk band edge energies with respect to the reference level and the vacuum energy. At last, the work function of 5.301 - 5.131 eV is predicted for Bi2Te3 (0001) surface and compared to various elements.
△ Less
Submitted 13 November, 2017;
originally announced November 2017.
-
Thermoelectric power factor of Bi-Sb-Te and Bi-Te-Se alloys and do** strategy: First-principles study
Authors:
Byungki Ryu,
Jaywan Chung,
Eun-Ae Choi,
Bong-Seo Kim,
Su-Dong Park
Abstract:
By performing first principles calculations combined with Boltzmann transport equations, we calculate the thermoelectric power factor (PF) of Bi-Sb-Te and Bi-Te-Se ternary alloys as a function of alloy composition ratio, carrier concentration, and temperature. The point defect formation energy calculations also perform to search potential n-type dopant candidates in ternaries.
By performing first principles calculations combined with Boltzmann transport equations, we calculate the thermoelectric power factor (PF) of Bi-Sb-Te and Bi-Te-Se ternary alloys as a function of alloy composition ratio, carrier concentration, and temperature. The point defect formation energy calculations also perform to search potential n-type dopant candidates in ternaries.
△ Less
Submitted 6 April, 2017;
originally announced April 2017.
-
Hybrid-Density Functional Theory Study on Band Structures of Tetradymite-Bi2Te3, Sb2Te3, Bi2Se3, and Sb2Se3 Thermoelectric Materials
Authors:
Sudong Park,
Byungki Ryu
Abstract:
The low energy band structure near the band gap determines the electrical performance of thermoelectric materials. Here, by using the hybrid-density functional theory (hybrid-DFT) calculations, we calculate the low energy band structure of Bi2Te3, Sb2Te3, Bi2Se3 and Sb2Se3 in tetradymite phase. We find that the band structure characteristics are very sensitive the selection of the exchange energy…
▽ More
The low energy band structure near the band gap determines the electrical performance of thermoelectric materials. Here, by using the hybrid-density functional theory (hybrid-DFT) calculations, we calculate the low energy band structure of Bi2Te3, Sb2Te3, Bi2Se3 and Sb2Se3 in tetradymite phase. We find that the band structure characteristics are very sensitive the selection of the exchange energy functional. The predictability of the band gaps and band degeneracies is not enhanced in hybrid-DFT calculations, as compared to DFT calculations. The poor prediction of low energy band structures is originated from the poor prediction of interlayer distances and the high structure sensitivity on the band gap. We conclude that the hybrid DFT calculations are not superior to DFT calculations when predicting band structures of tetradymite Bi2Te3, Sb2Te3, Bi2Se3 and Sb2Se3 thermoelectric materials.
△ Less
Submitted 20 October, 2016;
originally announced October 2016.
-
Prediction of Band Structure of $Bi_2Te_3$-related Binary and Ternary Thermoelectric Materials
Authors:
Byungki Ryu,
Min-Wook Oh,
Bong-Seo Kim,
Ji Eun Lee,
Sung-Jae Joo,
Bok-Ki Min,
HeeWoong Lee,
Sudong Park
Abstract:
Density functional calculations have performed to study the band structures of $Bi_2Te_3$-related binary ($Bi_2Te_3$, $Sb_2Te_3$, $Bi_2Se_3$, and $Sb_2Se_3$) and $Sb$/$Se$ doped ternary compounds [$(Bi_{1-x}Sb_x)_2Te_3$ and $Bi_2(Te_{1-y}Se_y)_3$]. It is found that the band gap can be increased by $Sb$ do** and it is monotonically increased by $Se$ do**. In ternary compounds, the change of the…
▽ More
Density functional calculations have performed to study the band structures of $Bi_2Te_3$-related binary ($Bi_2Te_3$, $Sb_2Te_3$, $Bi_2Se_3$, and $Sb_2Se_3$) and $Sb$/$Se$ doped ternary compounds [$(Bi_{1-x}Sb_x)_2Te_3$ and $Bi_2(Te_{1-y}Se_y)_3$]. It is found that the band gap can be increased by $Sb$ do** and it is monotonically increased by $Se$ do**. In ternary compounds, the change of the conduction band structure is more significant, as compared to the change of valence band. The band degeneracy of valence band maximum is maintained to be 6 in binaries and ternaries. However, as going from $Bi_2Te_3$ to $Sb_2Te_3$ ($Bi_2Se_3$), the degeneracy of conduction band minimum is reduced from 6 to 2(1). Based on the results of band structures, we suggest the suitable stoichiometries of ternary compounds for high thermoelectric performance.
△ Less
Submitted 14 November, 2016; v1 submitted 4 June, 2015;
originally announced June 2015.
-
Defects responsible for abnormal n-type conductivity in Ag-excess doped PbTe thermoelectrics
Authors:
Byungki Ryu,
Min-Wook Oh,
Jae Ki Lee,
Ji Eun Lee,
Sung-Jae Joo,
Bong-Seo Kim,
Bog-Ki Min,
Hee-Woong Lee,
Su-Dong Park
Abstract:
We find that Ag-interstitial ($Ag_I$) acts as an electron donor and plays an important role in Ag-excess doped polycrystalline PbTe thermoelectric materials. When Ag is heavily doped in PbTe, the neutral (Ag-Ag) dimer defect is formed at the Pb-site and the environment becomes Pb-rich/Te-poor condition. Then the positively ionized Ag interstitial ($Ag_I^+$) defect becomes the major defect under Pb…
▽ More
We find that Ag-interstitial ($Ag_I$) acts as an electron donor and plays an important role in Ag-excess doped polycrystalline PbTe thermoelectric materials. When Ag is heavily doped in PbTe, the neutral (Ag-Ag) dimer defect is formed at the Pb-site and the environment becomes Pb-rich/Te-poor condition. Then the positively ionized Ag interstitial ($Ag_I^+$) defect becomes the major defect under Pb-rich condition. Due to the small formation energy and small diffusion barrier of $Ag_I^+$, Ag can be easily dissolved into the PbTe matrix. The temperature behavior of the Ag defect formation energy well explains the $Ag_I^+$ solubility, the electron carrier generation, and the increasing electrical conductivity in Ag-excess doped polycrystalline PbTe at high temperature. This abnormal do** behavior by forming interstitial defects is also found for Au-doped PbTe.
△ Less
Submitted 4 June, 2015;
originally announced June 2015.
-
O-vacancy as the origin of negative bias illumination stress instability in amorphous In-Ga-Zn-O thin film transistors
Authors:
Byungki Ryu,
Hyeon-Kyun Noh,
Eun-Ae Choi,
K. J. Chang
Abstract:
We find that O-vacancy (Vo) acts as a hole trap and play a role in negative bias illumination stress instability in amorphous In-Ga-Zn-O thin film transistors. Photo-excited holes drifted toward the channel/dielectric interface due to small potential barriers and can be captured by Vo in the dielectrics. While Vo(+2) defects are very stable at room temperature, their original deep states are recov…
▽ More
We find that O-vacancy (Vo) acts as a hole trap and play a role in negative bias illumination stress instability in amorphous In-Ga-Zn-O thin film transistors. Photo-excited holes drifted toward the channel/dielectric interface due to small potential barriers and can be captured by Vo in the dielectrics. While Vo(+2) defects are very stable at room temperature, their original deep states are recovered via electron capture upon annealing. We also find that Vo(+2) can diffuse in amorphous phase, including hole accumulation near the interface under negative gate bias.
△ Less
Submitted 25 June, 2010;
originally announced June 2010.
