arXiv subjects

Yiheng Lin

Publications and source records attributed to Yiheng Lin.

67 records · Page 4Linked to original sources

arXiv2020

Scalable Multi-Agent Reinforcement Learning for Networked Systems with Average Reward

It has long been recognized that multi-agent reinforcement learning (MARL) faces significant scalability issues due to the fact that the size of the state and action spaces are exponentially large in the number of agents. In this paper, we identify a rich class of networked MARL problems where the model exhibits a local dependence structure that allows it to be solved in a scalable manner. Specifically, we propose a Scalable Actor-Critic (SAC) method that can learn a near optimal localized policy for optimizing the average reward with complexity scaling with the state-action space size of local neighborhoods, as opposed to the entire network. Our result centers around identifying and exploiting an exponential decay property that ensures the effect of agents on each other decays exponentially fast in their graph distance.

math.OC↗

arXiv2020

Quantum entanglement between an atom and a molecule

Conventional information processors freely convert information between different physical carriers to process, store, or transmit information. It seems plausible that quantum information will also be held by different physical carriers in applications such as tests of fundamental physics, quantum-enhanced sensors, and quantum information processing. Quantum-controlled molecules in particular could transduce quantum information across a wide range of quantum-bit (qubit) frequencies, from a few kHz for transitions within the same rotational manifold, a few GHz for hyperfine transitions, up to a few THz for rotational transitions, to hundreds of THz for fundamental and overtone vibrational and electronic transitions, possibly all within the same molecule. Here, we report the first demonstration of entanglement between states of the rotation of a $\rm^{40}CaH^+$ molecular ion and internal states of a $\rm^{40}Ca^+$ atomic ion. The qubit addressed in the molecule has a frequency of either 13.4 kHz or 855 GHz, highlighting the versatility of molecular qubits. This work demonstrates how molecules can transduce quantum information between qubits with different frequencies to enable hybrid quantum systems. We anticipate that quantum control and measurement of molecules as demonstrated here will create opportunities for quantum information science, quantum sensors, fundamental and applied physics, and controlled quantum chemistry.

quant-ph↗

arXiv2020

Online Optimization with Predictions and Non-convex Losses

We study online optimization in a setting where an online learner seeks to optimize a per-round hitting cost, which may be non-convex, while incurring a movement cost when changing actions between rounds. We ask: \textit{under what general conditions is it possible for an online learner to leverage predictions of future cost functions in order to achieve near-optimal costs?} Prior work has provided near-optimal online algorithms for specific combinations of assumptions about hitting and switching costs, but no general results are known. In this work, we give two general sufficient conditions that specify a relationship between the hitting and movement costs which guarantees that a new algorithm, Synchronized Fixed Horizon Control (SFHC), provides a $1+O(1/w)$ competitive ratio, where $w$ is the number of predictions available to the learner. Our conditions do not require the cost functions to be convex, and we also derive competitive ratio results for non-convex hitting and movement costs. Our results provide the first constant, dimension-free competitive ratio for online non-convex optimization with movement costs. Further, we give an example of a natural instance, Convex Body Chasing (CBC), where the sufficient conditions are not satisfied and we can prove that no online algorithm can have a competitive ratio that converges to 1.

cs.LG↗

arXiv2019

Precision frequency-comb terahertz spectroscopy on pure quantum states of a single molecular ion

Spectroscopy is a powerful tool for studying molecules and is commonly performed on large thermal molecular ensembles that are perturbed by motional shifts and interactions with the environment and one another, resulting in convoluted spectra and limited resolution. Here, we use generally applicable quantum-logic techniques to prepare a trapped molecular ion in a single quantum state, drive terahertz rotational transitions with an optical frequency comb, and read out the final state non-destructively, leaving the molecule ready for further manipulation. We resolve rotational transitions to 11 significant digits and derive the rotational constant of CaH+ to be B_R = 142501777.9(1.7) kHz. Our approach suits a wide range of molecular ions, including polyatomics and species relevant for tests of fundamental physics, chemistry, and astrophysics.

physics.atom-ph↗

arXiv2019

Beyond Online Balanced Descent: An Optimal Algorithm for Smoothed Online Optimization

We study online convex optimization in a setting where the learner seeks to minimize the sum of a per-round hitting cost and a movement cost which is incurred when changing decisions between rounds. We prove a new lower bound on the competitive ratio of any online algorithm in the setting where the costs are $m$-strongly convex and the movement costs are the squared $\ell_2$ norm. This lower bound shows that no algorithm can achieve a competitive ratio that is $o(m^{-1/2})$ as $m$ tends to zero. No existing algorithms have competitive ratios matching this bound, and we show that the state-of-the-art algorithm, Online Balanced Decent (OBD), has a competitive ratio that is $Ω(m^{-2/3})$. We additionally propose two new algorithms, Greedy OBD (G-OBD) and Regularized OBD (R-OBD) and prove that both algorithms have an $O(m^{-1/2})$ competitive ratio. The result for G-OBD holds when the hitting costs are quasiconvex and the movement costs are the squared $\ell_2$ norm, while the result for R-OBD holds when the hitting costs are $m$-strongly convex and the movement costs are Bregman Divergences. Further, we show that R-OBD simultaneously achieves constant, dimension-free competitive ratio and sublinear regret when hitting costs are strongly convex.

cs.LG↗

arXiv2019

Spin detection with a micromechanical trampoline: Towards magnetic resonance microscopy harnessing cavity optomechanics

We explore the prospects and benefits of combining the techniques of cavity optomechanics with efforts to image spins using magnetic resonance force microscopy (MRFM). In particular, we focus on a common mechanical resonator used in cavity optomechanics -- high-stress stoichiometric silicon nitride (Si$_3$N$_4$) membranes. We present experimental work with a trampoline membrane resonator that has a quality factor above $10^6$ and an order of magnitude lower mass than a comparable standard membrane resonators. Such high-stress resonators are on a trajectory to reach 0.1 $\rm{aN}/\sqrt{\rm{Hz}}$ force sensitivities at MHz frequencies by using techniques such as soft clamping and phononic-crystal control of acoustic radiation in combination with cryogenic cooling. We present a demonstration of force-detected electron spin resonance of an ensemble at room temperature using the trampoline resonators functionalized with a magnetic grain. We discuss prospects for combining such a resonator with an integrated Fabry-Perot cavity readout at cryogenic temperatures, and provide ideas for future impacts of membrane cavity optomechanical devices on MRFM of nuclear spins.

cond-mat.mes-hall↗

arXiv2018

Quantum optimal control of the dissipative production of a maximally entangled state

Entanglement generation can be robust against noise in approaches that deliberately incorporate dissipation into the system dynamics. The presence of additional dissipation channels may, however, limit fidelity and speed of the process. Here we show how quantum optimal control techniques can be used to both speed up the entanglement generation and increase the fidelity in a realistic setup, whilst respecting typical experimental limitations. For the example of entangling two trapped ion qubits [Lin et al., Nature 504, 415 (2013)], we find an improved fidelity by simply optimizing the polarization of the laser beams utilized in the experiment. More significantly, an alternate combination of transitions between internal states of the ions, when combined with optimized polarization, enables faster entanglement and decreases the error by an order of magnitude.

quant-ph↗

arXiv2018

Population mixing due to dipole-dipole interactions in a 1D array of multilevel atoms

We examine theoretically how dipole-dipole interactions arising from multiple photon scattering lead to a modified distribution of ground state populations in a driven, ordered 1D array of multilevel atoms. Specifically, we devise a level configuration in which a ground-state population accumulated due solely to dipole-dipole interactions can be up to 20\% in regimes accessible to current experiments with neutral atom arrays. For much larger systems, the steady state can consist of an equal distribution of population across the ground state manifold. Our results illustrate how dipole-dipole interactions can be accentuated through interference, and regulated by the geometry of ordered atom arrays. More generally, control techniques for multilevel atoms that can be degraded by multiple scattering, such as optical pumping, will benefit from an improved understanding and control of dipole-dipole interactions available in ordered arrays.

quant-ph↗

arXiv2018

Self-calibrating vector atomic magnetometry through microwave polarization reconstruction

Atomic magnetometry is one of the most sensitive ways to measure magnetic fields. We present a method for converting a naturally scalar atomic magnetometer into a vector magnetometer by exploiting the polarization dependence of hyperfine transitions in rubidium atoms. First, we fully determine the polarization ellipse of an applied microwave field using a self-calibrating method, i.e. a method in which the light-atom interaction provides everything required to know the field in an orthogonal laboratory frame. We then measure the direction of an applied static field using the polarization ellipse as a three-dimensional reference defined by Maxwell's equations. Although demonstrated with trapped atoms, this technique could be applied to atomic vapors, or a variety of atom-like systems.

physics.atom-ph↗

arXiv2018

Measurement-based entanglement of noninteracting bosonic atoms

We demonstrate the ability to extract a spin-entangled state of two neutral atoms via postselection based on a measurement of their spatial configuration. Typically, entangled states of neutral atoms are engineered via atom-atom interactions. In contrast, in our work we use Hong-Ou-Mandel interference to postselect a spin-singlet state after overlapping two atoms in distinct spin states on an effective beam splitter. We verify the presence of entanglement and determine a bound on the postselected fidelity of a spin-singlet state of $\left(0.62 \pm 0.03\right)$. The experiment has direct analogy to creating polarization entanglement with single photons and hence demonstrates the potential to use protocols developed for photons to create complex quantum states with noninteracting atoms.

physics.atom-ph↗

arXiv2017

Monolithic Integration of AlGaAs Distributed Bragg Reflectors on Virtual Ge Substrates via Aspect Ratio Trapping

High quality AlxGa1-xAs distributed Bragg reflectors (DBRs) were successfully monolithically grown on on-axis Si (100) substrates via a Ge layer formed by aspect ratio trapping (ART) technique. The GaAs/ART-Ge/Si-based DBRs have reflectivity spectra comparable to those grown on conventional bulk off-cut GaAs substrates and have smooth morphology, and good periodicity and uniformity. Anitphase domain formation is significantly reduced in GaAs on ART-Ge/Si substrates, and etch pit density of the GaAs base layer on the ART-Ge substrates ranges from 10^5 to 6 x 10^6 cm^(-2). These results paved the way for future VCSEL growth and fabrication on these ART-Ge substrates and also confirm that virtual Ge substrates via ART technique are effective Si platforms for optoelectronic integrated circuits.

cond-mat.mtrl-sci↗

arXiv2016

On the Origins of Near-Surface Stresses in Silicon around Cu-filled and CNT-filled Through Silicon Vias

Micro-Raman spectroscopy was employed to study the near-surface stress distributions and origins in Si around through silicon vias (TSVs) at both room temperature and elevated temperatures for Cu-filled and CNT-filled TSV samples. From the observations, we proved that the stresses near TSVs are mainly from two sources: 1) pre-existing stress before via filling, and 2) coefficients of thermal expansion (CTE) mismatch-induced stress. CTE-mismatch-induced stress is shown to dominate the compressive regime of the near-surface stress distribution around Cu-filled TSV structures, while pre-existing stress dominates the full range of the stress distribution in the CNT-filled TSV structures. Once the pre-existing stress is minimized, the total stress around CNT-filled TSVs can be minimized accordingly. Therefore, compared to Cu-filled TSVs, CNT-filled TSVs hold the potential to circumvent the hassle of stress-aware circuit layout and to solve the stress-related reliability issues.

cond-mat.mtrl-sci↗

arXiv2015

Dissipative quantum control of a spin chain

A protocol is discussed for preparing a spin chain in a generic many-body state in the asymptotic limit of tailored non-unitary dynamics. The dynamics require the spectral resolution of the target state, optimized coherent pulses, engineered dissipation, and feedback. As an example, we discuss the preparation of an entangled antiferromagnetic state, and argue that the procedure can be applied to chains of trapped ions or Rydberg atoms.

quant-ph↗