Search

We study the phase diagram and critical properties of quantum Ising chains with long-range ferromagnetic interactions decaying in a power-law fashion with exponent $\alpha$, in regimes of direct interest for current trapped ion experiments. Using large-scale path integral Monte Carlo simulations, we investigate both the ground-state and the nonzero-temperature regimes. We identify the phase boundary of the ferromagnetic phase and obtain accurate estimates for the ferromagnetic-paramagnetic transition temperatures. We further determine the critical exponents of the respective transitions. Our results are in agreement with existing predictions for interaction exponents $\alpha > 1$ up to small deviations in some critical exponents. We also address the elusive regime $\alpha < 1$, where we find that the universality class of both the ground-state and nonzero-temperature transition is consistent with the mean-field limit at $\alpha = 0$. Our work not only contributes to the understanding of the equilibrium properties of long-range interacting quantum Ising models, but can also be important for addressing fundamental dynamical aspects, such as issues concerning the open question of thermalization in such models., Comment: 19 pages, 6 figures, updated to follow minor revisions suggested by the referees

The increasing size of neural network models has been critical for improvements in their accuracy, but device memory is not growing at the same rate. This creates fundamental challenges for training neural networks within limited memory environments. In this work, we propose ActNN, a memory-efficient training framework that stores randomly quantized activations for back propagation. We prove the convergence of ActNN for general network architectures, and we characterize the impact of quantization on the convergence via an exact expression for the gradient variance. Using our theory, we propose novel mixed-precision quantization strategies that exploit the activation's heterogeneity across feature dimensions, samples, and layers. These techniques can be readily applied to existing dynamic graph frameworks, such as PyTorch, simply by substituting the layers. We evaluate ActNN on mainstream computer vision models for classification, detection, and segmentation tasks. On all these tasks, ActNN compresses the activation to 2 bits on average, with negligible accuracy loss. ActNN reduces the memory footprint of the activation by 12x, and it enables training with a 6.6x to 14x larger batch size., Comment: to be published in ICML 2021

The results of joint analyses of available cosmological data have motivated an important debate about a possible detection of a non-zero spatial curvature. If confirmed, such a result would imply a change in our present understanding of cosmic evolution with important theoretical and observational consequences. In this paper we discuss the legitimacy of carrying out joint analyses with the currently available data sets and explore their implications for a non-flat universe and extensions of the standard cosmological model. We use a robust tension estimator to perform a quantitative analysis of the physical consistency between the latest data of Cosmic Microwave Background, type Ia supernovae, Baryonic Acoustic Oscillations and Cosmic Chronometers. We consider the flat and non-flat cases of the $\Lambda$CDM cosmology and of two dark energy models with a constant and varying dark energy EoS parameter. The present study allows us to better understand if possible inconsistencies between these data sets are significant enough to make the results of their joint analyses misleading, as well as the actual dependence of such results with the spatial curvature and dark energy parameterizations., Comment: 11 pages, 6 figures, 1 table

Humans have a remarkable ability to make decisions by accurately reasoning about future events, including the future behaviors and states of mind of other agents. Consider driving a car through a busy intersection: it is necessary to reason about the physics of the vehicle, the intentions of other drivers, and their beliefs about your own intentions. If you signal a turn, another driver might yield to you, or if you enter the passing lane, another driver might decelerate to give you room to merge in front. Competent drivers must plan how they can safely react to a variety of potential future behaviors of other agents before they make their next move. This requires contingency planning: explicitly planning a set of conditional actions that depend on the stochastic outcome of future events. In this work, we develop a general-purpose contingency planner that is learned end-to-end using high-dimensional scene observations and low-dimensional behavioral observations. We use a conditional autoregressive flow model to create a compact contingency planning space, and show how this model can tractably learn contingencies from behavioral observations. We developed a closed-loop control benchmark of realistic multi-agent scenarios in a driving simulator (CARLA), on which we compare our method to various noncontingent methods that reason about multi-agent future behavior, including several state-of-the-art deep learning-based planning approaches. We illustrate that these noncontingent planning methods fundamentally fail on this benchmark, and find that our deep contingency planning method achieves significantly superior performance. Code to run our benchmark and reproduce our results is available at https://sites.google.com/view/contingency-planning, Comment: To be published at ICRA 2021. Project page: https://sites.google.com/view/contingency-planning

Traditional computer vision models are trained to predict a fixed set of predefined categories. Recently, natural language has been shown to be a broader and richer source of supervision that provides finer descriptions to visual concepts than supervised "gold" labels. Previous works, such as CLIP, use a simple pretraining task of predicting the pairings between images and text captions. CLIP, however, is data hungry and requires more than 400M image text pairs for training. We propose a data-efficient contrastive distillation method that uses soft labels to learn from noisy image-text pairs. Our model transfers knowledge from pretrained image and sentence encoders and achieves strong performance with only 3M image text pairs, 133x smaller than CLIP. Our method exceeds the previous SoTA of general zero-shot learning on ImageNet 21k+1k by 73% relatively with a ResNet50 image encoder and DeCLUTR text encoder. We also beat CLIP by 10.5% relatively on zero-shot evaluation on Google Open Images (19,958 classes)., Comment: 4 pages, 1 figure

Building reliable object detectors that are robust to domain shifts, such as various changes in context, viewpoint, and object appearances, is critical for real-world applications. In this work, we study the effectiveness of auxiliary self-supervised tasks to improve the out-of-distribution generalization of object detectors. Inspired by the principle of maximum entropy, we introduce a novel self-supervised task, instance-level temporal cycle confusion (CycConf), which operates on the region features of the object detectors. For each object, the task is to find the most different object proposals in the adjacent frame in a video and then cycle back to itself for self-supervision. CycConf encourages the object detector to explore invariant structures across instances under various motions, which leads to improved model robustness in unseen domains at test time. We observe consistent out-of-domain performance improvements when training object detectors in tandem with self-supervised tasks on large-scale video datasets (BDD100K and Waymo open data). The joint training framework also establishes a new state-of-the-art on standard unsupervised domain adaptative detection benchmarks (Cityscapes, Foggy Cityscapes, and Sim10K). The code and models are available at https://github.com/xinw1012/cycle-confusion., Comment: ICCV 2021

We analyze high-resolution (400pc) 220GHz continuum and CO(2-1) ALMA observations of a representative sample of 23 local (z<0.165) ULIRG systems (34 individual nuclei) as part of the "Physics of ULIRGs with MUSE and ALMA" (PUMA) project. The deconvolved half-light radii of the 220GHz continuum sources are between <60-350 pc (median 90pc). We associate these regions with the regions emitting the bulk of the infrared luminosity. The good agreement, within a factor of 2, between the 220GHz fluxes and the extrapolation of the infrared gray-body, and the small synchrotron and free-free contributions support this assumption. The cold molecular gas emission sizes, r_CO, are 60-700 pc and are similar in advanced mergers and early interacting systems. On average, r_CO are 2.5 times larger than the continuum. We derive L_IR and cold molecular gas surface densities: log Sigma(L_IR)=11.5-14.3 Lsun/kpc^2 and log Sigma(H2)=2.9-4.2 Msun/pc^2. Assuming that the L_IR is produced by star-formation, this corresponds to median Sigma(SFR)=2500 Msun/yr/kpc^2 which would imply extremely short depletion times, <1-15 Myr, and unphysical SF efficiencies >1 for 70% of the sample. Therefore, this favors the presence of obscured AGN that could dominate the L_IR. We also classify the ULIRG nuclei in two groups: (a) compact nuclei (r<130 pc) with high mid-IR excess emission found in optically classified AGN; and (b) nuclei following a relation with decreasing mid-IR excess for decreasing r. 60% of the interacting nuclei lie in the low end (<130 pc) of this relation, while only 30% of the advanced mergers do so, suggesting that in the early interaction phases the activity occurs in more compact and obscured regions. About two thirds of the nuclei are above the Eddington limit which is consistent with the detection of massive outflows in local ULIRGs and the potential role of radiation pressure in the launching process., Comment: Accepted for publication in A&A

We present Pylot, a platform for autonomous vehicle (AV) research and development, built with the goal to allow researchers to study the effects of the latency and accuracy of their models and algorithms on the end-to-end driving behavior of an AV. This is achieved through a modular structure enabled by our high-performance dataflow system that represents AV software pipeline components (object detectors, motion planners, etc.) as a dataflow graph of operators which communicate on data streams using timestamped messages. Pylot readily interfaces with popular AV simulators like CARLA, and is easily deployable to real-world vehicles with minimal code changes. To reduce the burden of developing an entire pipeline for evaluating a single component, Pylot provides several state-of-the-art reference implementations for the various components of an AV pipeline. Using these reference implementations, a Pylot-based AV pipeline is able to drive a real vehicle, and attains a high score on the CARLA Autonomous Driving Challenge. We also present several case studies enabled by Pylot, including evidence of a need for context-dependent components, and per-component time allocation. Pylot is open source, with the code available at https://github.com/erdos-project/pylot., Comment: ICRA 2021; 8 pages, 4 figures, 1 table

Context. The high number of super-Earth and Earth-like planets in the habitable zone (HZ) detected around M-dwarf stars in the last years has revealed these stellar objects to be the key for planetary radial velocity (RV) searches. Aims. Using the HARPS-N spectrograph within The HArps-n red Dwarf Exoplanet Survey (HADES) we reach the precision needed to detect small planets with a few Earth masses using the RV technique. Methods. We obtained 138 HARPS-N RV measurements between 2013 May and 2020 September of GJ 720 A, classified as an M0.5V star located at a distance of 15.56 pc. To characterize the stellar variability and to discern the periodic variation due to the Keplerian signals from those related to stellar activity, the HARPS-N spectroscopic activity indicators and the simultaneous photometric observations were analyzed. The combined analysis of HARPS-N RVs and activity indicators let us to address the nature of the periodic signals. The final model and the orbital planetary parameters were obtained by fitting simultaneously the stellar variability and the Keplerian signal using a Gaussian process regression and following a Bayesian criterion. Results. The HARPS-N RV periodic signals around 40 d and 100 d have counterparts at the same frequencies in HARPS-N activity indicators and photometric light curves. Then we attribute these periodicities to stellar activity the former period being likely associated with the stellar rotation. GJ 720 A shows the most significant signal at 19.466$\pm$0.005 d with no counterparts in any stellar activity indices. We hence ascribe this RV signal, having a semiamplitude of 4.72$\pm$0.27 m/s , to the presence of a sub-Neptune mass planet. The planet GJ 720 Ab has a minimum mass of 13.64$\pm$0.79 M$_{\oplus}$, it is in circular orbit at 0.119$\pm$0.002 AU from its parent star, and lies inside the inner boundary of the HZ around its parent star.

We propose opportunistic evaluation, a framework for accelerating interactions with dataframes. Interactive latency is critical for iterative, human-in-the-loop dataframe workloads for supporting exploratory data analysis. Opportunistic evaluation significantly reduces interactive latency by 1) prioritizing computation directly relevant to the interactions and 2) leveraging think time for asynchronous background computation for non-critical operators that might be relevant to future interactions. We show, through empirical analysis, that current user behavior presents ample opportunities for optimization, and the solutions we propose effectively harness such opportunities.

Dynamical histories of planetary systems, as well as atmospheric evolution of highly irradiated planets, can be studied by characterizing the ultra-short-period planet population, which the TESS mission is particularly well suited to discover. Here, we report on the follow-up of a transit signal detected in the TESS sector 19 photometric time series of the M3.0 V star TOI-1685 (2MASS J04342248+4302148). We confirm the planetary nature of the transit signal, which has a period of P_b=0.6691403+0.0000023-0.0000021 d, using precise radial velocity measurements taken with the CARMENES spectrograph. From the joint photometry and radial velocity analysis, we estimate the following parameters for TOI-1685 b: a mass of M_b=3.78+/-0.63 M_Earth, a radius of R_b=1.70+/-0.07 R_Earth, which together result in a bulk density of rho_b=4.21+0.95-0.82 g/cm3, and an equilibrium temperature of Teq_b=1069+/-16 K. TOI-1685 b is the least dense ultra-short period planet around an M dwarf known to date. TOI-1685 b is also one of the hottest transiting Earth-size planets with accurate dynamical mass measurements, which makes it a particularly attractive target for thermal emission spectroscopy. Additionally, we report a further non-transiting planet candidate in the system, TOI-1685[c], with an orbital period of P_[c]=9.02+0.10-0.12 d., Comment: 18 pages, 15 figures

Some luminous and ultraluminous infrared galaxies (LIRGs and ULIRGs) host extremely compact and dusty nuclei. The intense infrared radiation arising from warm dust in these sources is prone to excite vibrational levels of molecules such as HCN. This results in emission from the rotational transitions of vibrationally excited HCN (HCN-vib), with the brightest emission found in compact obscured nuclei (CONs). We aim to establish how common CONs are in the local Universe, and whether their prevalence depends on the luminosity or other properties of the host galaxy. We have conducted an Atacama Large Millimeter/submillimeter Array (ALMA) survey of the rotational J=3-2 transition of HCN-vib in a sample of 46 far-infrared luminous galaxies. Compact obscured nuclei are identified in 38 percent of ULIRGs, 21 percent of LIRGs, and 0 percent of lower luminosity galaxies. We find no dependence on the inclination of the host galaxy, but strong evidence of lower IRAS 25 to 60 {\mu}m flux density ratios (f25/f60) in CONs compared to the rest of the sample. Furthermore, we find that CONs have stronger silicate features (s9.7{\mu}m) but similar PAH equivalent widths (EQW6.2{\mu}m) compared to other galaxies. In the local Universe, CONs are primarily found in (U)LIRGs. High resolution continuum observations of the individual nuclei are required to determine if the CON phenomenon is related to the inclinations of the nuclear disks. The lower f25/f60 ratios in CONs as well as the results for the mid-infrared diagnostics investigated are consistent with large dust columns shifting the nuclear radiation to longer wavelengths, making the mid- and far-infrared "photospheres" significantly cooler than the interior regions. To assess the importance of CONs in the context of galaxy evolution, it is necessary to extend this study to higher redshifts where (U)LIRGs are more common., Comment: 16 pages, 11 figures, Accepted for publication in Astronomy and Astrophysics

M-dwarfs have proven to be ideal targets for planetary radial velocity (RV) searches due to their higher planet-star mass contrast. The HADES and CARMENES programs aim to carry out extensive searches of exoplanetary systems around this type of stars in the northern hemisphere, allowing us to address statistically the properties of the planets orbiting these objects. In this work, we perform a spectroscopic and photometric study of one of the program stars (GJ 740), which exhibits a short-period RV signal compatible with a planetary companion. We carried out a spectroscopic analysis based on 129 HARPS-N spectra taken over a time-span of 6 yr combined with 57 HARPS spectra taken over 4 yr, as well as 32 CARMENES spectra taken during more than 1 yr, resulting in a dataset with a time coverage of 10 yr. We also relied on 459 measurements from the public ASAS survey with a time-coverage of 8 yr along with 5 yr of photometric magnitudes from the EXORAP project taken in the $V$, $B$, $R$, and $I$ filters to carry out a photometric study. Both analyses were made using Markov Chain Monte Carlo (MCMC) simulations and Gaussian Process regression to model the activity of the star. We present the discovery of a short-period super-Earth with an orbital period of 2.37756$^{+0.00013}_{-0.00011}$ d and a minimum mass of 2.96$^{+0.50}_{-0.48}$ M$_{\oplus}$. We offer an update to the previously reported characterization of the magnetic cycle and rotation period of the star, obtaining values of $P_{\rm rot}$=35.563$\pm$0.071 d and $P_{\rm cycle}$=2800$\pm$150 d. Furthermore, the RV time-series exhibits a possibly periodic long-term signal which might be related to a Saturn-mass planet of $\sim$ 100 M$_{\oplus}$., Comment: 18 pages, 12 figures

i-TED is an innovative detection system which exploits Compton imaging techniques to achieve a superior signal-to-background ratio in ($n,\gamma$) cross-section measurements using time-of-flight technique. This work presents the first experimental validation of the i-TED apparatus for high-resolution time-of-flight experiments and demonstrates for the first time the concept proposed for background rejection. To this aim both $^{197}$Au($n,\gamma$) and $^{56}$Fe($n, \gamma$) reactions were measured at CERN n\_TOF using an i-TED demonstrator based on only three position-sensitive detectors. Two \cds detectors were also used to benchmark the performance of i-TED. The i-TED prototype built for this study shows a factor of $\sim$3 higher detection sensitivity than state-of-the-art \cds detectors in the $\sim$10~keV neutron energy range of astrophysical interest. This paper explores also the perspectives of further enhancement in performance attainable with the final i-TED array consisting of twenty position-sensitive detectors and new analysis methodologies based on Machine-Learning techniques., Comment: 16 pages, 16 figures

We describe mechanisms for the allocation of a scarce resource among multiple users in a way that is efficient, fair, and strategy-proof, but when users do not know their resource requirements. The mechanism is repeated for multiple rounds and a user's requirements can change on each round. At the end of each round, users provide feedback about the allocation they received, enabling the mechanism to learn user preferences over time. Such situations are common in the shared usage of a compute cluster among many users in an organisation, where all teams may not precisely know the amount of resources needed to execute their jobs. By understating their requirements, users will receive less than they need and consequently not achieve their goals. By overstating them, they may siphon away precious resources that could be useful to others in the organisation. We formalise this task of online learning in fair division via notions of efficiency, fairness, and strategy-proofness applicable to this setting, and study this problem under three types of feedback: when the users' observations are deterministic, when they are stochastic and follow a parametric model, and when they are stochastic and nonparametric. We derive mechanisms inspired by the classical max-min fairness procedure that achieve these requisites, and quantify the extent to which they are achieved via asymptotic rates. We corroborate these insights with an experimental evaluation on synthetic problems and a web-serving task.

Efficient exploration under sparse rewards remains a key challenge in deep reinforcement learning. To guide exploration, previous work makes extensive use of intrinsic reward (IR). There are many heuristics for IR, including visitation counts, curiosity, and state-difference. In this paper, we analyze the pros and cons of each method and propose the regulated difference of inverse visitation counts as a simple but effective criterion for IR. The criterion helps the agent explore Beyond the Boundary of explored regions and mitigates common issues in count-based methods, such as short-sightedness and detachment. The resulting method, BeBold, solves the 12 most challenging procedurally-generated tasks in MiniGrid with just 120M environment steps, without any curriculum learning. In comparison, the previous SoTA only solves 50% of the tasks. BeBold also achieves SoTA on multiple tasks in NetHack, a popular rogue-like game that contains more challenging procedurally-generated environments.

Researchers and practitioners in the field of reinforcement learning (RL) frequently leverage parallel computation, which has led to a plethora of new algorithms and systems in the last few years. In this paper, we re-examine the challenges posed by distributed RL and try to view it through the lens of an old idea: distributed dataflow. We show that viewing RL as a dataflow problem leads to highly composable and performant implementations. We propose RLlib Flow, a hybrid actor-dataflow programming model for distributed RL, and validate its practicality by porting the full suite of algorithms in RLlib, a widely adopted distributed RL library. Concretely, RLlib Flow provides 2-9 code savings in real production code and enables the composition of multi-agent algorithms not possible by end users before. The open-source code is available as part of RLlib at https://github.com/ray-project/ray/tree/master/rllib., Comment: NeurIPS 2021. The first two authors contributed equally to this work

Relaxing the conventional assumption of a minimal coupling between the dark matter (DM) and dark energy (DE) fields introduces significant changes in the predicted evolution of the Universe. Therefore, testing such a possibility constitutes an essential task not only for cosmology but also for fundamental physics. In a previous communication [Phys. Rev. D99, 043521, 2019], we proposed a new null test for the $\Lambda$CDM model based on the time dependence of the ratio between the DM and DE energy densities which is also able to detect potential signatures of interaction between the dark components. In this work, we extend that analysis avoiding the $ \Lambda$CDM assumption and reconstruct the interaction in the dark sector in a fully model-independent way using data from type Ia supernovae, cosmic chronometers and baryonic acoustic oscillations. According to our analysis, the $\Lambda$CDM model is consistent with our model-independent approach at least at $3\sigma$ CL over the entire range of redshift studied. On the other hand, our analysis shows that the current background data do not allow us to rule out the existence of an interaction in the dark sector. Finally, we present a forecast for next-generation LSS surveys. In particular, we show that Euclid and SKA will be able to distinguish interacting models with about 4\% of precision at $z\approx 1$., Comment: 12 pages, 7 figures. Accepted for publication in Physical Review D

Automation of surgical tasks using cable-driven robots is challenging due to backlash, hysteresis, and cable tension, and these issues are exacerbated as surgical instruments must often be changed during an operation. In this work, we propose a framework for automation of high-precision surgical tasks by learning sample efficient, accurate, closed-loop policies that operate directly on visual feedback instead of robot encoder estimates. This framework, which we call intermittent visual servoing (IVS), intermittently switches to a learned visual servo policy for high-precision segments of repetitive surgical tasks while relying on a coarse open-loop policy for the segments where precision is not necessary. To compensate for cable-related effects, we apply imitation learning to rapidly train a policy that maps images of the workspace and instrument from a top-down RGB camera to small corrective motions. We train the policy using only 180 human demonstrations that are roughly 2 seconds each. Results on a da Vinci Research Kit suggest that combining the coarse policy with half a second of corrections from the learned policy during each high-precision segment improves the success rate on the Fundamentals of Laparoscopic Surgery peg transfer task from 72.9% to 99.2%, 31.3% to 99.2%, and 47.2% to 100.0% for 3 instruments with differing cable-related effects. In the contexts we studied, IVS attains the highest published success rates for automated surgical peg transfer and is significantly more reliable than previous techniques when instruments are changed. Supplementary material is available at https://tinyurl.com/ivs-icra., Comment: 6 pages, 5 figures, 4 tables, submitted to ICRA 2021, supplementary material at https://tinyurl.com/ivs-icra

Untangling ropes, wires, and cables is a challenging task for robots due to the high-dimensional configuration space, visual homogeneity, self-occlusions, and complex dynamics. We consider dense (tight) knots that lack space between self-intersections and present an iterative approach that uses learned geometric structure in configurations. We instantiate this into an algorithm, HULK: Hierarchical Untangling from Learned Keypoints, which combines learning-based perception with a geometric planner into a policy that guides a bilateral robot to untangle knots. To evaluate the policy, we perform experiments both in a novel simulation environment modelling cables with varied knot types and textures and in a physical system using the da Vinci surgical robot. We find that HULK is able to untangle cables with dense figure-eight and overhand knots and generalize to varied textures and appearances. We compare two variants of HULK to three baselines and observe that HULK achieves 43.3% higher success rates on a physical system compared to the next best baseline. HULK successfully untangles a cable from a dense initial configuration containing up to two overhand and figure-eight knots in 97.9% of 378 simulation experiments with an average of 12.1 actions per trial. In physical experiments, HULK achieves 61.7% untangling success, averaging 8.48 actions per trial. Supplementary material, code, and videos can be found at https://tinyurl.com/y3a88ycu., Comment: Conference on Robot Learning (CoRL) 2020 Oral. First two authors contributed equally