Showing 1–20 of 20 results for author: Zhi, J

PATIENT-Ψ: Using Large Language Models to Simulate Patients for Training Mental Health Professionals

Authors: Ruiyi Wang, Stephanie Milani, Jamie C. Chiu, Jiayin Zhi, Shaun M. Eack, Travis Labrum, Samuel M. Murphy, Nev Jones, Kate Hardy, Hong Shen, Fei Fang, Zhiyu Zoey Chen

Abstract: Mental illness remains one of the most critical public health issues. Despite its importance, many mental health professionals highlight a disconnect between their training and actual real-world patient practice. To help bridge this gap, we propose PATIENT-Ψ, a novel patient simulation framework for cognitive behavior therapy (CBT) training. To build PATIENT-Ψ, we construct diverse patient cogniti… ▽ More Mental illness remains one of the most critical public health issues. Despite its importance, many mental health professionals highlight a disconnect between their training and actual real-world patient practice. To help bridge this gap, we propose PATIENT-Ψ, a novel patient simulation framework for cognitive behavior therapy (CBT) training. To build PATIENT-Ψ, we construct diverse patient cognitive models based on CBT principles and use large language models (LLMs) programmed with these cognitive models to act as a simulated therapy patient. We propose an interactive training scheme, PATIENT-Ψ-TRAINER, for mental health trainees to practice a key skill in CBT -- formulating the cognitive model of the patient -- through role-playing a therapy session with PATIENT-Ψ. To evaluate PATIENT-Ψ, we conducted a comprehensive user study of 13 mental health trainees and 20 experts. The results demonstrate that practice using PATIENT-Ψ-TRAINER enhances the perceived skill acquisition and confidence of the trainees beyond existing forms of training such as textbooks, videos, and role-play with non-patients. Based on the experts' perceptions, PATIENT-Ψ is perceived to be closer to real patient interactions than GPT-4, and PATIENT-Ψ-TRAINER holds strong promise to improve trainee competencies. Our code and data are released at \url{https://github.com/ruiyiw/patient-psi}. △ Less

Submitted 18 June, 2024; v1 submitted 29 May, 2024; originally announced May 2024.

Comments: 9 pages, 5 figures

Statistical analysis of pulsar flux density distribution

Authors: H. W. Xu, R. S. Zhao, Erbil Gugercinoglu, H. Liu, D. Li, P. Wang, C. H. Niu, C. Miao, X. Zhu, R. W. Tian, W. L. Li, S. D. Wang, Z. F. Tu, Q. J. Zhi, S. J. Dang, L. H. Shang, S. Xiao

Abstract: This study presents a comprehensive analysis of the spectral properties of 886 pulsars across a wide frequency range from 20MHz to 343.5GHz, including a total of 86 millisecond pulsars. The majority of the pulsars exhibit power-law behavior in their spectra, although some exceptions are observed. Five different spectral models, namely simple power-law, broken power-law, low-frequency turn-over, hi… ▽ More This study presents a comprehensive analysis of the spectral properties of 886 pulsars across a wide frequency range from 20MHz to 343.5GHz, including a total of 86 millisecond pulsars. The majority of the pulsars exhibit power-law behavior in their spectra, although some exceptions are observed. Five different spectral models, namely simple power-law, broken power-law, low-frequency turn-over, high-frequency cut-off, and double turn-over, were employed to explore the spectral behaviors. The average spectral index for pulsars modeled with a simple power-law is found to be -1.64 +/-0.80, consistent with previous studies. Additionally, significant correlations between the spectral index and characteristic parameters are observed particularly in millisecond pulsars, while no strong correlation is observed in normal pulsars. Different models show variations in the most influential characteristic parameters associated with the spectral index, indicating diverse dominant radiation mechanisms in millisecond pulsars.Finally, this study identifies 22 pulsars of the Gigahertz-peaked Spectra (GPS) type for the first time based on the Akaike information criterion. △ Less

Submitted 16 April, 2024; v1 submitted 16 April, 2024; originally announced April 2024.

Comments: 39 papers,17figures

Ultrafast and precise distance measurement via real-time chirped pulse interferometry

Authors: Mingyang Xu, Hanzhong Wu, Jiawen Zhi, Yang Liu, Jie Zhang, Zehuang Lu, Chenggang Shao

Abstract: Laser frequency combs, which are composed of a series of equally-spaced coherent frequency components, have triggered revolutionary progress for precision spectroscopy and optical metrology. Length/distance is of fundamental importance in both science and technology. In this work, we describe a ranging scheme based on chirped pulse interferometry. In contrast to the traditional spectral interferom… ▽ More Laser frequency combs, which are composed of a series of equally-spaced coherent frequency components, have triggered revolutionary progress for precision spectroscopy and optical metrology. Length/distance is of fundamental importance in both science and technology. In this work, we describe a ranging scheme based on chirped pulse interferometry. In contrast to the traditional spectral interferometry, the local oscillator is strongly chirped which is able to meet the measurement pulses at arbitrary distances, and therefore the dead zones can be removed. The distances can be precisely determined via two measurement steps based on time-of-flight method and synthetic wavelength interferometry, respectively. To overcome the speed limitation of the optical spectrum analyzer, the spectrograms are stretched and detected by a fast photodetector and oscilloscope, and consequently mapped into the time domain in real time. The experimental results indicate that the measurement uncertainty can be well within 2 $\upmu$m, compared with the reference distance meter. The Allan deviation can reach 0.4 $\upmu$m at averaging time of 4 ns, 25 nm at 1 $\upmu$s, and can achieve 2 nm at 100 $\upmu$s averaging time. We also measure a spinning disk with grooves of different depths to verify the measurement speed, and the results show that the grooves with about 150 m/s line speed can be clearly captured. Our method provides a unique combination of non-dead zones, ultrafast measurement speed, high precision and accuracy, large ambiguity range, and with only one single comb source. This system could offer a powerful solution for the field measurements in practical applications in future. △ Less

Submitted 25 February, 2024; originally announced February 2024.

Discovery of four pulsars in a pilot survey at intermediate Galactic latitudes with FAST

Authors: Q. J. Zhi, J. T. Bai, S. Dai, X. Xu, S. J. Dang, L. H. Shang, R. S. Zhao, D. Li, W. W. Zhu, N. Wang, J. P. Yuan, P. Wang, L. Zhang, Y. Feng, J. B. Wang, S. Q. Wang, Q. D. Wu, A. J. Dong, H. Yang, J. Tian, W. Q. Zhong, X. H. Luo, Miroslav D. Filipovi, G. J. Qiao

Abstract: We present the discovery and timing results of four pulsars discovered in a pilot survey at intermediate Galactic latitudes with the Five-hundred Aperture Spherical Telescope (FAST). Among these pulsars, two belong to the category of millisecond pulsars (MSPs) with spin periods of less than 20 ms. The other two fall under the classification of "mildly recycled" pulsars, with massive white dwarfs a… ▽ More We present the discovery and timing results of four pulsars discovered in a pilot survey at intermediate Galactic latitudes with the Five-hundred Aperture Spherical Telescope (FAST). Among these pulsars, two belong to the category of millisecond pulsars (MSPs) with spin periods of less than 20 ms. The other two fall under the classification of "mildly recycled" pulsars, with massive white dwarfs as companions. Remarkably, this small survey, covering an area of 4.7 $deg^2$ , led to the discovery of four recycled pulsars. Such success underscores the immense potential of future surveys at intermediate Galactic latitudes. In order to assess the potential yield of MSPs, we conducted population simulations and found that both FAST and Parkes new phased array feed surveys, focusing on intermediate Galactic latitudes, have the capacity to uncover several hundred new MSPs. △ Less

Submitted 28 December, 2023; v1 submitted 1 November, 2023; originally announced November 2023.

Comments: 7 pages, 4 figures, 2 tables, accepted to ApJ

Experimental demonstration of picometer level signal extraction with time-delay interferometry technique

Authors: Mingyang Xu, Yujie Tan, Yurong Liang, Jiawen Zhi, Xiaoyang Guo, Dan Luo, Panpan Wang, Hanzhong Wu, Chenggang Shao

Abstract: In this work, we have built an experimental setup to simulate the clock noise transmission with two spacecrafts and two optical links, and further demonstrated the extraction of picometer level signal drowned by the large laser frequency noise and clock noise with the data post-processing method. Laser frequency noise is almost eliminated by using the idea of time-delay interferometry (TDI) to con… ▽ More In this work, we have built an experimental setup to simulate the clock noise transmission with two spacecrafts and two optical links, and further demonstrated the extraction of picometer level signal drowned by the large laser frequency noise and clock noise with the data post-processing method. Laser frequency noise is almost eliminated by using the idea of time-delay interferometry (TDI) to construct an equal arm interferometer. Clock asynchronism and clock jitter noise are significantly suppressed by laser sideband transmitting the clock noise using an electro-optic modulator (EOM). Experimental results show a reduction in laser frequency noise by approximately 10^5 and clock noise by 10^2, recovering a weak displacement signal with an average amplitude about 60 picometer and period 1 second. This work has achieved the principle verification of the noise reduction function of TDI technique to some extent, serving the data processing research of space-borne gravitational wave detection. △ Less

Submitted 26 October, 2023; originally announced October 2023.

Improving Human-Robot Collaboration via Computational Design

Authors: Jixuan Zhi, Jyh-Ming Lien

Abstract: When robots entered our day-to-day life, the shared space surrounding humans and robots is critical for effective Human-Robot collaboration. The design of shared space should satisfy humans' preferences and robots' efficiency. This work uses kitchen design as an example to illustrate the importance of good space design in facilitating such collaboration. Given the kitchen boundary, counters, and r… ▽ More When robots entered our day-to-day life, the shared space surrounding humans and robots is critical for effective Human-Robot collaboration. The design of shared space should satisfy humans' preferences and robots' efficiency. This work uses kitchen design as an example to illustrate the importance of good space design in facilitating such collaboration. Given the kitchen boundary, counters, and recipes, the proposed method computes the optimal placement of counters that meet the requirement of kitchen design rules and improve Human-Robot collaboration. The key technical challenge is that the optimization method usually evaluates thousands of designs and the computational cost of motion planning, which is part of the evaluation function, is expensive. We use a decentralized motion planner that can solve multi-agent motion planning efficiently. Our results indicate that optimized kitchen designs can provide noticeable performance improvement to Human-Robot collaboration. △ Less

Submitted 20 March, 2023; originally announced March 2023.

Decomposing User-APP Graph into Subgraphs for Effective APP and User Embedding Learning

Authors: Tan Yu, Jun Zhi, Yufei Zhang, Jian Li, Hongliang Fei, Ping Li

Abstract: APP-installation information is helpful to describe the user's characteristics. The users with similar APPs installed might share several common interests and behave similarly in some scenarios. In this work, we learn a user embedding vector based on each user's APP-installation information. Since the user APP-installation embedding is learnable without dependency on the historical intra-APP behav… ▽ More APP-installation information is helpful to describe the user's characteristics. The users with similar APPs installed might share several common interests and behave similarly in some scenarios. In this work, we learn a user embedding vector based on each user's APP-installation information. Since the user APP-installation embedding is learnable without dependency on the historical intra-APP behavioral data of the user, it complements the intra-APP embedding learned within each specific APP. Thus, they considerably help improve the effectiveness of the personalized advertising in each APP, and they are particularly beneficial for the cold start of the new users in the APP. In this paper, we formulate the APP-installation user embedding learning into a bipartite graph embedding problem. The main challenge in learning an effective APP-installation user embedding is the imbalanced data distribution. In this case, graph learning tends to be dominated by the popular APPs, which billions of users have installed. In other words, some niche/specialized APPs might have a marginal influence on graph learning. To effectively exploit the valuable information from the niche APPs, we decompose the APP-installation graph into a set of subgraphs. Each subgraph contains only one APP node and the users who install the APP. For each mini-batch, we only sample the users from the same subgraph in the training process. Thus, each APP can be involved in the training process in a more balanced manner. After integrating the learned APP-installation user embedding into our online personal advertising platform, we obtained a considerable boost in CTR, CVR, and revenue. △ Less

Submitted 13 October, 2022; originally announced October 2022.

Detection of strong scattering close to the eclipse region of PSR B1957+20

Authors: J. T. Bai, S. Dai, Q. J. Zhi, W. A. Coles, D. Li, W. W. Zhu, G. Hobbs, G. J. Qiao, N. Wang, J. P. Yuan, M. D. Filipovic, J. B. Wang, Z. C. Pan, L. H. Shang, S. J. Dang, S. Q. Wang, C. C. Miao

Abstract: We present the first measurement of pulse scattering close to the eclipse region of PSR B1957+20, which is in a compact binary system with a low-mass star. We measured pulse scattering time-scales up to 0.2 ms close to the eclipse and showed that it scales with the dispersion measure (DM) excess roughly as $τ\proptoΔ{\rm DM}^{2}$. Our observations provide the first evidence of strong scattering du… ▽ More We present the first measurement of pulse scattering close to the eclipse region of PSR B1957+20, which is in a compact binary system with a low-mass star. We measured pulse scattering time-scales up to 0.2 ms close to the eclipse and showed that it scales with the dispersion measure (DM) excess roughly as $τ\proptoΔ{\rm DM}^{2}$. Our observations provide the first evidence of strong scattering due to multi-path propagation effects in the eclipsing material. We show that Kolmogorov turbulence in the eclipsing material with an inner scale of $\sim100$ m and an outer scale of the size of the eclipse region can naturally explain the observation. Our results show that the eclipsing material in such systems can be highly turbulent and suggest that scattering is one of the main eclipsing mechanisms at around 1.4 GHz. △ Less

Submitted 28 March, 2022; originally announced March 2022.

Comments: 8 pages, 4 figures, MNRAS accepted

A bimodal burst energy distribution of a repeating fast radio burst source

Authors: D. Li, P. Wang, W. W. Zhu, B. Zhang, X. X. Zhang, R. Duan, Y. K. Zhang, Y. Feng, N. Y. Tang, S. Chatterjee, J. M. Cordes, M. Cruces, S. Dai, V. Gajjar, G. Hobbs, C. Jin, M. Kramer, D. R. Lorimer, C. C. Miao, C. H. Niu, J. R. Niu, Z. C. Pan, L. Qian, L. Spitler, D. Werthimer , et al. (7 additional authors not shown)

Abstract: The event rate, energy distribution, and time-domain behaviour of repeating fast radio bursts (FRBs) contains essential information regarding their physical nature and central engine, which are as yet unknown. As the first precisely-localized source, FRB 121102 has been extensively observed and shows non-Poisson clustering of bursts over time and a power-law energy distribution. However, the exten… ▽ More The event rate, energy distribution, and time-domain behaviour of repeating fast radio bursts (FRBs) contains essential information regarding their physical nature and central engine, which are as yet unknown. As the first precisely-localized source, FRB 121102 has been extensively observed and shows non-Poisson clustering of bursts over time and a power-law energy distribution. However, the extent of the energy distribution towards the fainter end was not known. Here we report the detection of 1652 independent bursts with a peak burst rate of 122~hr^{-1}, in 59.5 hours spanning 47 days. A peak in the isotropic equivalent energy distribution is found to be ~4.8 x 10^{37} erg at 1.25~GHz, below which the detection of bursts is suppressed. The burst energy distribution is bimodal, and well characterized by a combination of a log-normal function and a generalized Cauchy function. The large number of bursts in hour-long spans allow sensitive periodicity searches between 1 ms and 1000 s. The non-detection of any periodicity or quasi-periodicity poses challenges for models involving a single rotating compact object. The high burst rate also implies that FRBs must be generated with a high radiative efficiency, disfavoring emission mechanisms with large energy requirements or contrived triggering conditions. △ Less

Submitted 14 October, 2021; v1 submitted 17 July, 2021; originally announced July 2021.

Comments: 69 pages, 14 figures, 2 tables; Updates for author correction

Journal ref: Nature 598 (2021) 267-271

Designing Human-Robot Coexistence Space

Authors: Jixuan Zhi, Lap-Fai Yu, Jyh-Ming Lien

Abstract: When the human-robot interactions become ubiquitous, the environment surrounding these interactions will have significant impact on the safety and comfort of the human and the effectiveness and efficiency of the robot. Although most robots are designed to work in the spaces created for humans, many environments, such as living rooms and offices, can be and should be redesigned to enhance and impro… ▽ More When the human-robot interactions become ubiquitous, the environment surrounding these interactions will have significant impact on the safety and comfort of the human and the effectiveness and efficiency of the robot. Although most robots are designed to work in the spaces created for humans, many environments, such as living rooms and offices, can be and should be redesigned to enhance and improve human-robot collaboration and interactions. This work uses autonomous wheelchair as an example and investigates the computational design in the human-robot coexistence spaces. Given the room size and the objects $O$ in the room, the proposed framework computes the optimal layouts of $O$ that satisfy both human preferences and navigation constraints of the wheelchair. The key enabling technique is a motion planner that can efficiently evaluate hundreds of similar motion planning problems. Our implementation shows that the proposed framework can produce a design around three to five minutes on average comparing to 10 to 20 minutes without the proposed motion planner. Our results also show that the proposed method produces reasonable designs even for tight spaces and for users with different preferences. △ Less

Submitted 14 November, 2020; originally announced November 2020.

Luminosity of radio pulsar and its new emission death line

Authors: Q. D. Wu, Q. J. Zhi, C. M. Zhang, D. H. Wang, C. Q. Ye

Abstract: We investigated the pulsar radio luminosity ($L$), emission efficiency (ratio of radio luminosity to its spin-down power $\dot{E}$), and death line in the diagram of magnetic field (B) versus spin period (P), and found that the dependence of pulsar radio luminosity on its spin-down power ($L-\dot{E}$) is very weak, shown as $L\sim\dot{E}^{0.06}$, which deduces an equivalent inverse correlation bet… ▽ More We investigated the pulsar radio luminosity ($L$), emission efficiency (ratio of radio luminosity to its spin-down power $\dot{E}$), and death line in the diagram of magnetic field (B) versus spin period (P), and found that the dependence of pulsar radio luminosity on its spin-down power ($L-\dot{E}$) is very weak, shown as $L\sim\dot{E}^{0.06}$, which deduces an equivalent inverse correlation between emission efficiency and spin-down power as $ξ\sim \dot{E}^{-0.94}$. Furthermore, we examined the distributions of radio luminosity of millisecond and normal pulsars, and found that, for the similar spin-down powers, the radio luminosity of millisecond pulsars is about one order of magnitude lower than that of the normal pulsars. The analysis of pulsar radio flux suggests that this correlations are not due to a selective effect, but are intrinsic to the pulsar radio emission physics. Their radio radiations may be dominated by the different radiation mechanisms. The cut-off phenomenon of currently observed radio pulsars in B-P diagram is usually referred as the "pulsar death line", which corresponds to $\dot{E}\approx 10^{30}$ erg/s and is obtained by the cut-off voltage of electron acceleration gap in the polar cap model of pulsar proposed by Ruderman and Sutherland. Observationally, this death line can be inferred by the actual observed pulsar flux $S\approx $1mJy and 1kpc distance, together with the maximum radio emission efficiency of 1\%. At present, the actual observed pulsar flux can reach 0.01mJy by FAST telescope, which will arise the observational limit of spin-down power of pulsar as low as $\dot{E}\approx 10^28$ erg/s. This means that the new death line is downward shifted two orders of magnitude, which might be favorably referred as the "observational limit-line", and accordingly the pulsar theoretical model for the cut-off voltage of gap should be heavily modified. △ Less

Submitted 2 July, 2020; originally announced July 2020.

Learning to Herd Agents Amongst Obstacles: Training Robust Shepherding Behaviors using Deep Reinforcement Learning

Authors: Jixuan Zhi, Jyh-Ming Lien

Abstract: Robotic shepherding problem considers the control and navigation of a group of coherent agents (e.g., a flock of bird or a fleet of drones) through the motion of an external robot, called shepherd. Machine learning based methods have successfully solved this problem in an empty environment with no obstacles. Rule-based methods, on the other hand, can handle more complex scenarios in which environm… ▽ More Robotic shepherding problem considers the control and navigation of a group of coherent agents (e.g., a flock of bird or a fleet of drones) through the motion of an external robot, called shepherd. Machine learning based methods have successfully solved this problem in an empty environment with no obstacles. Rule-based methods, on the other hand, can handle more complex scenarios in which environments are cluttered with obstacles and allow multiple shepherds to work collaboratively. However, these rule-based methods are fragile due to the difficulty in defining a comprehensive set of rules that can handle all possible cases. To overcome these limitations, we propose the first known learning-based method that can herd agents amongst obstacles. By using deep reinforcement learning techniques combined with the probabilistic roadmaps, we train a shepherding model using noisy but controlled environmental and behavioral parameters. Our experimental results show that the proposed method is robust, namely, it is insensitive to the uncertainties originated from both environmental and behavioral models. Consequently, the proposed method has a higher success rate, shorter completion time and path length than the rule-based behavioral methods have. These advantages are particularly prominent in more challenging scenarios involving more difficult groups and strenuous passages. △ Less

Submitted 19 May, 2020; originally announced May 2020.

An in-depth investigation of 11 pulsars discovered by FAST

Authors: A. D. Cameron, D. Li, G. Hobbs, L. Zhang, C. C. Miao, J. B. Wang, M. Yuan, S. Wang, G. Jacobs Corban, M. Cruces, S. Dai, Y. Feng, J. Han, J. F. Kaczmarek, J. R. Nui, Z. C. Pan, L. Qian, Z. Z. Tao, P. Wang, S. Q. Wang, H. Xu, R. X. Xu, Y. L. Yue, S. B. Zhang, Q. J. Zhi , et al. (6 additional authors not shown)

Abstract: We present timing solutions and analyses of 11 pulsars discovered by the Five-hundred-meter Aperture Spherical radio Telescope (FAST). These pulsars were discovered using an ultra-wide bandwidth receiver in drift-scan observations made during the commissioning phase of FAST, and were then confirmed and timed using the 64-m Parkes Radio Telescope. Each pulsar has been observed over a span of at lea… ▽ More We present timing solutions and analyses of 11 pulsars discovered by the Five-hundred-meter Aperture Spherical radio Telescope (FAST). These pulsars were discovered using an ultra-wide bandwidth receiver in drift-scan observations made during the commissioning phase of FAST, and were then confirmed and timed using the 64-m Parkes Radio Telescope. Each pulsar has been observed over a span of at least one year. Highlighted discoveries include PSR J0344-0901, which displays mode-changing behaviour and may belong to the class of so-called `swooshing' pulsars (alongside PSRs B0919+06 and B1859+07); PSR J0803-0942, whose emission is almost completely linearly polarised; and PSRs J1900-0134 and J1945+1211, whose well defined polarisation angle curves place stringent constraints on their emission geometry. We further discuss the detectability of these pulsars by earlier surveys, and highlight lessons learned from our work in carrying out confirmation and monitoring observations of pulsars discovered by a highly sensitive telescope, many of which may be applicable to next-generation pulsar surveys. This paper marks one of the first major releases of FAST-discovered pulsars, and paves the way for future discoveries anticipated from the Commensal Radio Astronomy FAST Survey (CRAFTS). △ Less

Submitted 31 May, 2020; v1 submitted 18 May, 2020; originally announced May 2020.

Comments: 18 pages, 8 figures, 5 tables. Accepted for publication by MNRAS. Correcting error in the caption of Figure 8

Fiber: A Platform for Efficient Development and Distributed Training for Reinforcement Learning and Population-Based Methods

Authors: Jiale Zhi, Rui Wang, Jeff Clune, Kenneth O. Stanley

Abstract: Recent advances in machine learning are consistently enabled by increasing amounts of computation. Reinforcement learning (RL) and population-based methods in particular pose unique challenges for efficiency and flexibility to the underlying distributed computing frameworks. These challenges include frequent interaction with simulations, the need for dynamic scaling, and the need for a user interf… ▽ More Recent advances in machine learning are consistently enabled by increasing amounts of computation. Reinforcement learning (RL) and population-based methods in particular pose unique challenges for efficiency and flexibility to the underlying distributed computing frameworks. These challenges include frequent interaction with simulations, the need for dynamic scaling, and the need for a user interface with low adoption cost and consistency across different backends. In this paper we address these challenges while still retaining development efficiency and flexibility for both research and practical applications by introducing Fiber, a scalable distributed computing framework for RL and population-based methods. Fiber aims to significantly expand the accessibility of large-scale parallel computation to users of otherwise complicated RL and population-based approaches without the need to for specialized computational expertise. △ Less

Submitted 24 March, 2020; originally announced March 2020.

Enhanced POET: Open-Ended Reinforcement Learning through Unbounded Invention of Learning Challenges and their Solutions

Authors: Rui Wang, Joel Lehman, Aditya Rawal, Jiale Zhi, Yulun Li, Jeff Clune, Kenneth O. Stanley

Abstract: Creating open-ended algorithms, which generate their own never-ending stream of novel and appropriately challenging learning opportunities, could help to automate and accelerate progress in machine learning. A recent step in this direction is the Paired Open-Ended Trailblazer (POET), an algorithm that generates and solves its own challenges, and allows solutions to goal-switch between challenges t… ▽ More Creating open-ended algorithms, which generate their own never-ending stream of novel and appropriately challenging learning opportunities, could help to automate and accelerate progress in machine learning. A recent step in this direction is the Paired Open-Ended Trailblazer (POET), an algorithm that generates and solves its own challenges, and allows solutions to goal-switch between challenges to avoid local optima. However, the original POET was unable to demonstrate its full creative potential because of limitations of the algorithm itself and because of external issues including a limited problem space and lack of a universal progress measure. Importantly, both limitations pose impediments not only for POET, but for the pursuit of open-endedness in general. Here we introduce and empirically validate two new innovations to the original algorithm, as well as two external innovations designed to help elucidate its full potential. Together, these four advances enable the most open-ended algorithmic demonstration to date. The algorithmic innovations are (1) a domain-general measure of how meaningfully novel new challenges are, enabling the system to potentially create and solve interesting challenges endlessly, and (2) an efficient heuristic for determining when agents should goal-switch from one problem to another (helping open-ended search better scale). Outside the algorithm itself, to enable a more definitive demonstration of open-endedness, we introduce (3) a novel, more flexible way to encode environmental challenges, and (4) a generic measure of the extent to which a system continues to exhibit open-ended innovation. Enhanced POET produces a diverse range of sophisticated behaviors that solve a wide range of environmental challenges, many of which cannot be solved through other means. △ Less

Submitted 13 April, 2020; v1 submitted 18 March, 2020; originally announced March 2020.

Comments: 23 pages, 14 figures

An Atari Model Zoo for Analyzing, Visualizing, and Comparing Deep Reinforcement Learning Agents

Authors: Felipe Petroski Such, Vashisht Madhavan, Rosanne Liu, Rui Wang, Pablo Samuel Castro, Yulun Li, Jiale Zhi, Ludwig Schubert, Marc G. Bellemare, Jeff Clune, Joel Lehman

Abstract: Much human and computational effort has aimed to improve how deep reinforcement learning algorithms perform on benchmarks such as the Atari Learning Environment. Comparatively less effort has focused on understanding what has been learned by such methods, and investigating and comparing the representations learned by different families of reinforcement learning (RL) algorithms. Sources of friction… ▽ More Much human and computational effort has aimed to improve how deep reinforcement learning algorithms perform on benchmarks such as the Atari Learning Environment. Comparatively less effort has focused on understanding what has been learned by such methods, and investigating and comparing the representations learned by different families of reinforcement learning (RL) algorithms. Sources of friction include the onerous computational requirements, and general logistical and architectural complications for running Deep RL algorithms at scale. We lessen this friction, by (1) training several algorithms at scale and releasing trained models, (2) integrating with a previous Deep RL model release, and (3) releasing code that makes it easy for anyone to load, visualize, and analyze such models. This paper introduces the Atari Zoo framework, which contains models trained across benchmark Atari games, in an easy-to-use format, as well as code that implements common modes of analysis and connects such models to a popular neural network visualization library. Further, to demonstrate the potential of this dataset and software package, we show initial quantitative and qualitative comparisons between the performance and representations of several deep RL algorithms, highlighting interesting and previously unknown distinctions between them. △ Less

Submitted 29 May, 2019; v1 submitted 17 December, 2018; originally announced December 2018.

DE CVn: an eclipsing post-common envelope binary with a circumbinary disk and a giant planet

Authors: Z. T Han, S. B. Qian, L. Y. Zhu, Q. J. Zhi, A. J. Dong, B. Soonthornthum, S. Poshyachinda, T. Sarotsakulchai, X. H. Fang, Q. S. Wang, Irina Voloshina

Abstract: We present a timing analysis of the eclipsing post-common envelope binary (PCEB) DE CVn. Based on new CCD photometric observations and the published data, we found that the orbital period in DE CVn has a cyclic period oscillation with an amplitude of $28.08$ s and a period of $11.22$ years plus a rapid period decrease at a rate of $\dot{P}=-3.35\times10^{-11}ss^{-1}$. According to the evolutionary… ▽ More We present a timing analysis of the eclipsing post-common envelope binary (PCEB) DE CVn. Based on new CCD photometric observations and the published data, we found that the orbital period in DE CVn has a cyclic period oscillation with an amplitude of $28.08$ s and a period of $11.22$ years plus a rapid period decrease at a rate of $\dot{P}=-3.35\times10^{-11}ss^{-1}$. According to the evolutionary theory, secular period decreases in PCEBs arise from angular momentum losses (AMLs) driven by gravitational radiation (GR) and magnetic braking (MB). However, the observed orbital decay is too fast to be produced by AMLs via GR and MB, indicating that there could be other AML mechanism. We suggest that a circumbinary disk around DE CVn may be responsible for the additional AML. The disk mass was derived as a few$\times$$10^{-4}$-$10^{-3}$$M_{\odot}$ , which is in agreement with that inferred from previous studies in the order of magnitude. The cyclic change is most likely result of the gravitational perturbation by a circumbinary object due to the Applegate's mechanism fails to explain such a large period oscillation. The mass of the potential third body is calculated as $M_{3}\sin{i'}=0.011(\pm0.003)M_{\odot}$. Supposing the circumbinary companion and the eclipsing binary is coplanar, its mass would correspond to a giant planet. This hypothetical giant planet is moving in a circular orbit of radius $\sim5.75(\pm2.02)$ AU around its host star. △ Less

Submitted 21 November, 2018; originally announced November 2018.

Comments: 19 pages, 5 figures

Journal ref: 2018, ApJ, 868, 53

Investigating multi-frequency pulse profiles of PSRs B0329+54 and B1642-03 in an inverse Compton scattering (ICS) model

Authors: L. H. Shang, J. G. Lu, Y. J. Du, L. F. Hao, D. Li, K. J. Lee, Bin Li, L. X. Li, G. J. Qiao, Z. Q. Shen, D. H. Wang, M. Wang, X. J. Wu, Y. J. Wu, R. X. Xu, Y. L. Yue, Z. Yan, Q. J. Zhi, R. B. Zhao, R. S. Zhao

Abstract: The emission geometries, e.g. the emission region height, the beam shape, and radius-to-frequency mapping, are important predictions of pulsar radiation model. The multi-band radio observations carry such valuable information. In this paper, we study two bright pulsars, (PSRs B0329+54 and B1642-03) and observe them in high frequency (2.5 GHz, 5 GHz, and 8 GHz). The newly acquired data together wit… ▽ More The emission geometries, e.g. the emission region height, the beam shape, and radius-to-frequency mapping, are important predictions of pulsar radiation model. The multi-band radio observations carry such valuable information. In this paper, we study two bright pulsars, (PSRs B0329+54 and B1642-03) and observe them in high frequency (2.5 GHz, 5 GHz, and 8 GHz). The newly acquired data together with historical archive provide an atlas of multi-frequency profiles spanning from 100 MHz to 10 GHz. We study the frequency evolution of pulse profiles and the radiation regions with the these data. We firstly fit the pulse profiles with Gaussian functions to determine the phase of each component, and then calculate the radiation altitudes of different emission components and the radiation regions. We find that the inverse Compton scattering (ICS) model can reproduce the radiation geometry of these two pulsars. But for PSR B0329+54 the radiation can be generated in either annular gap (AG) or core gap (CG), while the radiation of PSR B1642-03 can only be generated in the CG. This difference is caused by the inclination angle and the impact angle of these two pulsars. The relation of beaming angle (the angle between the radiation direction and the magnetic axis) and the radiation altitudes versus frequency is also presented by modelling the beam-frequency evolution in the ICS model. The multi-band pulse profiles of these two pulsars can be described well by the ICS model combined with the CG and AG. △ Less

Submitted 10 March, 2017; originally announced March 2017.

Probing the accretion disc structure by the twin kHz QPOs and spins of neutron stars in LMXBs

Authors: D. H. Wang, C. M. Zhang, Y. J. Lei, L. Chen, J. L. Qu, Q. J. Zhi

Abstract: We analyze the relation between the emission radii of twin kilohertz quasi-periodic oscillations (kHz QPOs) and the co-rotation radii of the 12 neutron star low mass X-ray binaries (NS-LMXBs) which are simultaneously detected with the twin kHz QPOs and NS spins. We find that the average co-rotation radius of these sources is r_co about 32 km, and all the emission positions of twin kHz QPOs lie ins… ▽ More We analyze the relation between the emission radii of twin kilohertz quasi-periodic oscillations (kHz QPOs) and the co-rotation radii of the 12 neutron star low mass X-ray binaries (NS-LMXBs) which are simultaneously detected with the twin kHz QPOs and NS spins. We find that the average co-rotation radius of these sources is r_co about 32 km, and all the emission positions of twin kHz QPOs lie inside the corotation radii, indicating that the twin kHz QPOs are formed in the spin-up process. It is noticed that the upper frequency of twin kHz QPOs is higher than NS spin frequency by > 10%, which may account for a critical velocity difference between the Keplerian motion of accretion matter and NS spin that is corresponding to the production of twin kHz QPOs. In addition, we also find that about 83% of twin kHz QPOs cluster around the radius range of 15-20 km, which may be affected by the hard surface or the local strong magnetic field of NS. As a special case, SAX J1808.4-3658 shows the larger emission radii of twin kHz QPOs of r about 21-24 km, which may be due to its low accretion rate or small measured NS mass (< 1.4 solar mass). △ Less

Submitted 9 January, 2017; originally announced January 2017.

Comments: 7 pages, 5 figures, 1 table

Free-Standing Two-Dimensional Single-Crystalline InSb Nanosheets

Authors: Dong Pan, Dingxun Fan, Ning Kang, Jinhua Zhi, Xuezhe Yu, Hongqi Xu, Jianhua Zhao

Abstract: Growth of high-quality single-crystalline InSb layers remains challenging in material science. Such layered InSb materials are highly desired for searching for and manipulation of Majorana fermions in solid state, a fundamental research task in physics today, and for development of novel high-speed nanoelectronic and infrared optoelectronic devices. Here we report on a new route towards growth of… ▽ More Growth of high-quality single-crystalline InSb layers remains challenging in material science. Such layered InSb materials are highly desired for searching for and manipulation of Majorana fermions in solid state, a fundamental research task in physics today, and for development of novel high-speed nanoelectronic and infrared optoelectronic devices. Here we report on a new route towards growth of single-crystalline, layered InSb materials. We demonstrate the successful growth of free-standing, two-dimensional InSb nanosheets on one-dimensional InAs nanowires by molecular-beam epitaxy. The grown InSb nanosheets are pure zinc-blende single crystals. The length and width of the InSb nanosheets are up to several micrometers and the thickness is down to ~10 nm. The InSb nanosheets show a clear ambipolar behavior and a high electron mobility. Our work will open up new technology routes towards the development of InSb-based devices for applications in nanoelectronics, optoelectronics and quantum electronics, and for study of fundamental physical phenomena. △ Less

Submitted 20 November, 2015; originally announced November 2015.

Comments: 32 pages,20 figures

Journal ref: Nano Letters, 16, 2016, 834-841