arXiv subjects

Jongmin Lee

Publications and source records attributed to Jongmin Lee.

At least 73 records · Page 4Linked to original sources

arXiv2022

COptiDICE: Offline Constrained Reinforcement Learning via Stationary Distribution Correction Estimation

We consider the offline constrained reinforcement learning (RL) problem, in which the agent aims to compute a policy that maximizes expected return while satisfying given cost constraints, learning only from a pre-collected dataset. This problem setting is appealing in many real-world scenarios, where direct interaction with the environment is costly or risky, and where the resulting policy should comply with safety constraints. However, it is challenging to compute a policy that guarantees satisfying the cost constraints in the offline RL setting, since the off-policy evaluation inherently has an estimation error. In this paper, we present an offline constrained RL algorithm that optimizes the policy in the space of the stationary distribution. Our algorithm, COptiDICE, directly estimates the stationary distribution corrections of the optimal policy with respect to returns, while constraining the cost upper bound, with the goal of yielding a cost-conservative policy for actual constraint satisfaction. Experimental results show that COptiDICE attains better policies in terms of constraint satisfaction and return-maximization, outperforming baseline algorithms.

cs.LG↗

arXiv2021

A Mølmer-Sørensen Gate with Rydberg-Dressed Atoms

Neutral atoms are building blocks of ground-up quantum many-body systems. Well-controlled and high-fidelity entangling gates are an essential component for realizing complex neutral atom architectures for quantum computing, quantum simulation, and measurement with precision better than the standard quantum limit. In this Letter we report the realization of a Mølmer-Sørensen unitary between two neutral atoms, based on adiabatic single-photon dressing to Rydberg levels. We show that this technique is highly robust to noise sources and experimental imperfections that have limited the fidelity of other approaches to neutral atom gates.

quant-ph↗

arXiv2021

A Geometric Structure of Acceleration and Its Role in Making Gradients Small Fast

Since Nesterov's seminal 1983 work, many accelerated first-order optimization methods have been proposed, but their analyses lacks a common unifying structure. In this work, we identify a geometric structure satisfied by a wide range of first-order accelerated methods. Using this geometric insight, we present several novel generalizations of accelerated methods. Most interesting among them is a method that reduces the squared gradient norm with $\mathcal{O}(1/K^4)$ rate in the prox-grad setup, faster than the $\mathcal{O}(1/K^3)$ rates of Nesterov's FGM or Kim and Fessler's FPGM-m.

math.OC↗

arXiv2021

A passively pumped vacuum package sustaining cold atoms for more than 200 days

Compact cold-atom sensors depend on vacuum technology. One of the major limitations to miniaturizing these sensors are the active pumps -- typically ion pumps -- required to sustain the low pressure needed for laser cooling. Although passively pumped chambers have been proposed as a solution to this problem, technical challenges have prevented successful operation at the levels needed for cold-atom experiments. We present the first demonstration of a vacuum package successfully independent of ion pumps for more than a week; our vacuum package is capable of sustaining a cloud of cold atoms in a magneto-optical trap (MOT) for greater than 200 days using only non-evaporable getters and a rubidium dispenser. Measurements of the MOT lifetime indicate the package maintains a pressure of better than $2\times10^{-7}$ Torr. This result will significantly impact the development of compact atomic sensors, including those sensitive to magnetic fields, where the absence of an ion pump will be advantageous.

physics.atom-ph↗

arXiv2021

OptiDICE: Offline Policy Optimization via Stationary Distribution Correction Estimation

We consider the offline reinforcement learning (RL) setting where the agent aims to optimize the policy solely from the data without further environment interactions. In offline RL, the distributional shift becomes the primary source of difficulty, which arises from the deviation of the target policy being optimized from the behavior policy used for data collection. This typically causes overestimation of action values, which poses severe problems for model-free algorithms that use bootstrapping. To mitigate the problem, prior offline RL algorithms often used sophisticated techniques that encourage underestimation of action values, which introduces an additional set of hyperparameters that need to be tuned properly. In this paper, we present an offline RL algorithm that prevents overestimation in a more principled way. Our algorithm, OptiDICE, directly estimates the stationary distribution corrections of the optimal policy and does not rely on policy-gradients, unlike previous offline RL algorithms. Using an extensive set of benchmark datasets for offline RL, we show that OptiDICE performs competitively with the state-of-the-art methods.

cs.LG↗

arXiv2021

Spherical Principal Curves

This paper presents a new approach for dimension reduction of data observed in a sphere. Several dimension reduction techniques have recently developed for the analysis of non-Euclidean data. As a pioneer work, Hauberg (2016) attempted to implement principal curves on Riemannian manifolds. However, this approach uses approximations to deal with data on Riemannian manifolds, which causes distorted results. In this study, we propose a new approach to construct principal curves on a sphere by a projection of the data onto a continuous curve. Our approach lies in the same line of Hastie and Stuetzle (1989) that proposed principal curves for Euclidean space data. We further investigate the stationarity of the proposed principal curves that satisfy the self-consistency on a sphere. Results from real data analysis with earthquake data and simulation examples demonstrate the promising empirical properties of the proposed approach.

stat.ME↗

arXiv2021

Demonstration of a MOT in a Sub-Millimeter Membrane Hole

We demonstrate the generation of a cold-atom ensemble within a sub-millimeter diameter hole in a transparent membrane, a so-called "membrane MOT". With a sub-Doppler cooling process, the atoms trapped by the membrane MOT are cooled down to 10 uK. The atom number inside the unbridged/bridged membrane hole is about 10^4 to 10^5, and the 1/e^2-diameter of the MOT cloud is about 180 um for a 400 um-diameter membrane hole. Such a membrane device can, in principle, efficiently load cold atoms into the evanescent-field optical trap generated by the suspended membrane waveguide for strong atom-light interaction and provide the capability of sufficient heat dissipation at the waveguide. This represents a key step toward the photonic atom trap integrated platform (ATIP).

quant-ph↗

arXiv2021

Characterization of Suspended Membrane Waveguides towards a Photonic Atom Trap Integrated Platform

We demonstrate an optical waveguide device, capable of supporting the high, in-vacuum, optical power necessary for trapping a single atom or a cold atom ensemble with evanescent fields. Our photonic integrated platforms, with suspended membrane waveguides, successfully manages optical powers of 6 mW (500 um span) to nearly 30 mW (125 um span) over an un-tethered waveguide span. This platform is compatible with laser cooling and magneto-optical traps (MOTs) in the vicinity of the suspended waveguide, called the membrane MOT and the needle MOT, a key ingredient for efficient trap loading. We evaluate two novel designs that explore critical thermal management features that enable this large power handling. This work represents a significant step toward an integrated platform for coupling neutral atom quantum systems to photonic and electronic integrated circuits on silicon.

quant-ph↗

arXiv2020

Learning to Compose Hypercolumns for Visual Correspondence

Feature representation plays a crucial role in visual correspondence, and recent methods for image matching resort to deeply stacked convolutional layers. These models, however, are both monolithic and static in the sense that they typically use a specific level of features, e.g., the output of the last layer, and adhere to it regardless of the images to match. In this work, we introduce a novel approach to visual correspondence that dynamically composes effective features by leveraging relevant layers conditioned on the images to match. Inspired by both multi-layer feature composition in object detection and adaptive inference architectures in classification, the proposed method, dubbed Dynamic Hyperpixel Flow, learns to compose hypercolumn features on the fly by selecting a small number of relevant layers from a deep convolutional neural network. We demonstrate the effectiveness on the task of semantic correspondence, i.e., establishing correspondences between images depicting different instances of the same object or scene category. Experiments on standard benchmarks show that the proposed method greatly improves matching performance over the state of the art in an adaptive and efficient manner.

cs.CV↗

arXiv2020

Highly ordered lead-free double perovskite halides by design

Lead-free double perovskite halides are emerging optoelectronic materials that are alternatives to lead-based perovskite halides. Recently, single-crystalline double perovskite halides were synthesized, and their intriguing functional properties were demonstrated. Despite such pioneering works, lead-free double perovskite halides with better crystallinity are still in demand for applications to novel optoelectronic devices. Here, we realized highly crystalline Cs2AgBiBr6 single crystals with a well-defined atomic ordering on the microscopic scale. We avoided the formation of Ag vacancies and the subsequent secondary Cs3Bi2Br9 by manipulating the initial chemical environments in hydrothermal synthesis. The suppression of Ag vacancies allows us to reduce the trap density in the as-grown crystals and to enhance the carrier mobility further. Our design strategy is applicable for fabricating other lead-free halide materials with high crystallinity.

cond-mat.mtrl-sci↗

arXiv2019

Artificially Engineered Nanostrain in Iron Chalcogenide Superconductor Thin Film for Enhancing Supercurrent

Although nanoscale deformation, such as nanostrain in iron chalcogenide (FeSexTe1-x, FST) thin films, has attracted attention owing to the enhancement of general superconducting properties, including critical current density (Jc) and critical transition temperature, its formation has proven to be an extremely challenging and complex process thus far. Herein, we successfully fabricated an epitaxial FST thin film with uniformly distributed nanostrain by injection of a trace amount of CeO2 inside FST matrix using sequential pulsed laser deposition. Using transmission electron microscopy and geometrical phase analysis, we verified that a trace amount of CeO2 injection forms nanoscale fine defects with a nanostrained region, which has a tensile strain (ezz ~ 0.02) along the c-axis of the FST matrix. The nanostrained FST thin film achieves a remarkable Jc of 3.5 MA/cm2 for a self-field at 6 K and a highly enhanced Jc under the entire magnetic field with respect to a pristine FST thin film.

cond-mat.supr-con↗

arXiv2019

SPair-71k: A Large-scale Benchmark for Semantic Correspondence

Establishing visual correspondences under large intra-class variations, which is often referred to as semantic correspondence or semantic matching, remains a challenging problem in computer vision. Despite its significance, however, most of the datasets for semantic correspondence are limited to a small amount of image pairs with similar viewpoints and scales. In this paper, we present a new large-scale benchmark dataset of semantically paired images, SPair-71k, which contains 70,958 image pairs with diverse variations in viewpoint and scale. Compared to previous datasets, it is significantly larger in number and contains more accurate and richer annotations. We believe this dataset will provide a reliable testbed to study the problem of semantic correspondence and will help to advance research in this area. We provide the results of recent methods on our new dataset as baselines for further research. Our benchmark is available online at http://cvlab.postech.ac.kr/research/SPair-71k/.

cs.CV↗

arXiv2019

Hyperpixel Flow: Semantic Correspondence with Multi-layer Neural Features

Establishing visual correspondences under large intra-class variations requires analyzing images at different levels, from features linked to semantics and context to local patterns, while being invariant to instance-specific details. To tackle these challenges, we represent images by "hyperpixels" that leverage a small number of relevant features selected among early to late layers of a convolutional neural network. Taking advantage of the condensed features of hyperpixels, we develop an effective real-time matching algorithm based on Hough geometric voting. The proposed method, hyperpixel flow, sets a new state of the art on three standard benchmarks as well as a new dataset, SPair-71k, which contains a significantly larger number of image pairs than existing datasets, with more accurate and richer annotations for in-depth analysis.

cs.CV↗

arXiv2018

Attentive Semantic Alignment with Offset-Aware Correlation Kernels

Semantic correspondence is the problem of establishing correspondences across images depicting different instances of the same object or scene class. One of recent approaches to this problem is to estimate parameters of a global transformation model that densely aligns one image to the other. Since an entire correlation map between all feature pairs across images is typically used to predict such a global transformation, noisy features from different backgrounds, clutter, and occlusion distract the predictor from correct estimation of the alignment. This is a challenging issue, in particular, in the problem of semantic correspondence where a large degree of image variations is often involved. In this paper, we introduce an attentive semantic alignment method that focuses on reliable correlations, filtering out distractors. For effective attention, we also propose an offset-aware correlation kernel that learns to capture translation-invariant local transformations in computing correlation values over spatial locations. Experiments demonstrate the effectiveness of the attentive model and offset-aware kernel, and the proposed model combining both techniques achieves the state-of-the-art performance.

cs.CV↗

arXiv2018

Observation of out-of-plane spin texture in a SrTiO3 (111) two-dimensional electron gas

We explore the second order bilinear magnetoelectric resistance (BMER) effect in the d-electron-based two-dimensional electron gas (2DEG) at the SrTiO3 (111) surface. We find an evidence of a spin-split band structure with the archetypal spin-momentum locking of the Rashba effect for the in-plane component. Under an out-of-plane magnetic field, we find a BMER signal that breaks the six-fold symmetry of the electronic dispersion, which is a fingerprint for the presence of a momentum dependent out-of-plane spin component. Relativistic electronic structure calculations reproduce this spin-texture and indicate that the out-of-plane component is a ubiquitous property of oxide 2DEGs arising from strong crystal field effects. We further show that the BMER response of the SrTiO3 (111) 2DEG is tunable and unexpectedly large.

cond-mat.mes-hall↗

arXiv2018

Consensus-based Distributed Quantile Estimation in Sensor Networks

A quantile is defined as a value below which random draws from a given distribution falls with a given probability. In a centralized setting where the cumulative distribution function (CDF) is unknown, the empirical CDF (ECDF) can be used to estimate such quantiles after aggregating the data. In a fully distributed sensor network, however, it is challenging to estimate quantiles. This is because each sensor node observes local measurement data with limited storage and data transmission power which make it difficult to obtain the global ECDF. This paper proposes consensus-based quantile estimation for such a distributed network. The states of the proposed algorithm are recursively updated with two-steps at each iteration: one is a local update based on the measurement data and the current state, and the other is averaging the updated states with neighboring nodes. We consider the realistic case of communication links between nodes being corrupted by independent random noise. It is shown that the estimated state sequence is asymptotically unbiased and converges toward the sample quantile in the mean-square sense. The two step-size sequences corresponding to the averaging and local update steps result in a mixed-time scale algorithm with proper decay rates in order to achieve convergence. We also provide applications to distributed estimation of trimmed mean, computation of median, maximum, or minimum values and identification of outliers through simulation.

eess.SY↗

arXiv2016

Arbitrary Dicke-State Control of Symmetric Rydberg Ensembles

Symmetric ensembles of neutral atoms interacting via the Rydberg blockade are well-described by the Jaynes-Cummings Hamiltonian. We use this framework to study the problem of generating arbitrary superpositions of Dicke states of hyperfine qubits in such ensembles. The combination of the symmetric Rydberg blockade and microwaves that drive the qubits with a time-dependent phase is sufficient to make these ensembles completely controllable, in the sense that one can generate an arbitrary unitary transformation on the system. We apply this to the problem of state mapping. With currently feasible parameters, it is possible to generate arbitrary symmetric states of ~ 10 hypefine qubits with high fidelity in ~ 1 microsecond, assuming fast microwave phase switching times. To reduce the requirements on phase switching, we propose a "dressed ground control" scheme, in which the control task is simplified by restricting the system's dynamics to the dressed ground subspace.

quant-ph↗

arXiv2016

Demonstration of the Jaynes-Cummings ladder with Rydberg-dressed atoms

We observe the nonlinearity of the Jaynes-Cummings (JC) ladder in the Autler-Townes spectroscopy of the hyperfine ground states for a Rydberg-dressed two-atom system. Here the role of the two-level system in the JC model is played by the presence or absence of a collective Rydberg excitation, and the bosonic mode manifests as the number $n$ of single atom spin flips, symmetrically distributed between the atoms. We measure the normal-mode splitting and $\sqrt{n}$ nonlinearity as a function of detuning and Rabi frequency, thereby experimentally establishing the isomorphism with the JC model.

quant-ph↗