Insert related papers here

Insert related papers here

Insert related papers here

Insert related papers here

Insert related papers here

Insert related papers here

Welcome to Byte Size Arxiv

Papers made digestable

2023-02-02

STEPS: Joint Self-supervised Nighttime Image Enhancement and Depth Estimation

  • Detailed ablations also reveal the mechanism of our proposal
  • It brings meaningful new challenges to the community
  • Codes, data, and models are available at https://github.com/ucaszyp/STEPS.
Self-supervised depth estimation draws a lot of attention recently as it can promote the 3D sensing capabilities of self-driving vehicles. However, it intrinsically relies upon the photometric consistency assumption, which hardly holds during nighttime. Although various supervised nighttime image enhancement methods have been proposed, their generalization performance in challenging driving scenarios is not satisfactory. To this end, we propose the first method that jointly learns a nighttime image enhancer and a depth estimator, without using ground truth for either task. Our method tightly entangles two self-supervised tasks using a newly proposed uncertain pixel masking strategy. This strategy originates from the observation that nighttime images not only suffer from underexposed regions but also from overexposed regions. By fitting a bridge-shaped curve to the illumination map distribution, both regions are suppressed and two tasks are bridged naturally. We benchmark the method on two established datasets: nuScenes and RobotCar and demonstrate state-of-the-art performance on both of them. Detailed ablations also reveal the mechanism of our proposal. Last but not least, to mitigate the problem of sparse ground truth of existing datasets, we provide a new photo-realistically enhanced nighttime dataset based upon CARLA. It brings meaningful new challenges to the community. Codes, data, and models are available at https://github.com/ucaszyp/STEPS.

Authors: Yupeng Zheng, Chengliang Zhong, Pengfei Li, Huan-ang Gao, Yuhang Zheng, Bu Jin, Ling Wang, Hao Zhao, Guyue Zhou, Qichao Zhang, Dongbin Zhao.

2023-02-02

Lower Bounds for Learning in Revealing POMDPs

  • We establish strong PAC and regret lower bounds for learning in revealing POMDPs
  • Technically, our hard instance construction adapts techniques in \emph{distribution testing}, which is new to the RL literature and may be of independent interest.
This paper studies the fundamental limits of reinforcement learning (RL) in the challenging \emph{partially observable} setting. While it is well-established that learning in Partially Observable Markov Decision Processes (POMDPs) requires exponentially many samples in the worst case, a surge of recent work shows that polynomial sample complexities are achievable under the \emph{revealing condition} -- A natural condition that requires the observables to reveal some information about the unobserved latent states. However, the fundamental limits for learning in revealing POMDPs are much less understood, with existing lower bounds being rather preliminary and having substantial gaps from the current best upper bounds. We establish strong PAC and regret lower bounds for learning in revealing POMDPs. Our lower bounds scale polynomially in all relevant problem parameters in a multiplicative fashion, and achieve significantly smaller gaps against the current best upper bounds, providing a solid starting point for future studies. In particular, for \emph{multi-step} revealing POMDPs, we show that (1) the latent state-space dependence is at least $\Omega(S^{1.5})$ in the PAC sample complexity, which is notably harder than the $\widetilde{\Theta}(S)$ scaling for fully-observable MDPs; (2) Any polynomial sublinear regret is at least $\Omega(T^{2/3})$, suggesting its fundamental difference from the \emph{single-step} case where $\widetilde{O}(\sqrt{T})$ regret is achievable. Technically, our hard instance construction adapts techniques in \emph{distribution testing}, which is new to the RL literature and may be of independent interest.

Authors: Fan Chen, Huan Wang, Caiming Xiong, Song Mei, Yu Bai.

2023-02-02

Bayesian Metric Learning for Uncertainty Quantification in Image Retrieval

  • We propose the first Bayesian encoder for metric learning
  • We actualize this by first proving that the contrastive loss is a valid log-posterior
  • We then propose three methods that ensure a positive definite Hessian
  • Lastly, we present a novel decomposition of the Generalized Gauss-Newton approximation.
We propose the first Bayesian encoder for metric learning. Rather than relying on neural amortization as done in prior works, we learn a distribution over the network weights with the Laplace Approximation. We actualize this by first proving that the contrastive loss is a valid log-posterior. We then propose three methods that ensure a positive definite Hessian. Lastly, we present a novel decomposition of the Generalized Gauss-Newton approximation. Empirically, we show that our Laplacian Metric Learner (LAM) estimates well-calibrated uncertainties, reliably detects out-of-distribution examples, and yields state-of-the-art predictive performance.

Authors: Frederik Warburg, Marco Miani, Silas Brack, Soren Hauberg.

2023-02-02

Evidence for suppression of structure growth in the concordance cosmological model

  • We present evidence for a suppressed growth rate of large-scale structure during the dark-energy dominated era
  • When combined, they yield $\gamma=0.633^{+0.025}_{-0.024}$, excluding $\gamma=0.55$ at a statistical significance of 3.7$\sigma$.
We present evidence for a suppressed growth rate of large-scale structure during the dark-energy dominated era. Modeling the growth rate of perturbations with the ``growth index'' $\gamma$, we find that current cosmological data strongly prefer a higher growth index than the value $\gamma=0.55$ predicted by general relativity in a flat $\Lambda$CDM cosmology. Both the cosmic microwave background data from Planck and the large-scale structure data from weak lensing, galaxy clustering, and cosmic velocities separately favor growth suppression. When combined, they yield $\gamma=0.633^{+0.025}_{-0.024}$, excluding $\gamma=0.55$ at a statistical significance of 3.7$\sigma$. The combination of $f\sigma_8$ and Planck measurements prefers an even higher growth index of $\gamma=0.639^{+0.024}_{-0.025}$, corresponding to a 4.2$\sigma$-tension with the concordance model. In Planck data, the suppressed growth rate offsets the preference for nonzero curvature and fits the data equally well as the latter model. A higher $\gamma$ leads to a higher matter fluctuation amplitude $S_8$ inferred from galaxy clustering and weak lensing measurements, and a lower $S_8$ from Planck data, effectively resolving the $S_8$ tension.

Authors: Nhat-Minh Nguyen, Dragan Huterer, Yuewei Wen.

2023-02-02

SceneDreamer: Unbounded 3D Scene Generation from 2D Image Collections

  • Our framework is learned from in-the-wild 2D image collections only, without any 3D annotations
  • The height field represents the surface elevation of 3D scenes, while the semantic field provides detailed scene semantics
  • Lastly, a neural volumetric renderer, learned from 2D image collections through adversarial training, is employed to produce photorealistic images
  • Extensive experiments demonstrate the effectiveness of SceneDreamer and superiority over state-of-the-art methods in generating vivid yet diverse unbounded 3D worlds.
In this work, we present SceneDreamer, an unconditional generative model for unbounded 3D scenes, which synthesizes large-scale 3D landscapes from random noises. Our framework is learned from in-the-wild 2D image collections only, without any 3D annotations. At the core of SceneDreamer is a principled learning paradigm comprising 1) an efficient yet expressive 3D scene representation, 2) a generative scene parameterization, and 3) an effective renderer that can leverage the knowledge from 2D images. Our framework starts from an efficient bird's-eye-view (BEV) representation generated from simplex noise, which consists of a height field and a semantic field. The height field represents the surface elevation of 3D scenes, while the semantic field provides detailed scene semantics. This BEV scene representation enables 1) representing a 3D scene with quadratic complexity, 2) disentangled geometry and semantics, and 3) efficient training. Furthermore, we propose a novel generative neural hash grid to parameterize the latent space given 3D positions and the scene semantics, which aims to encode generalizable features across scenes. Lastly, a neural volumetric renderer, learned from 2D image collections through adversarial training, is employed to produce photorealistic images. Extensive experiments demonstrate the effectiveness of SceneDreamer and superiority over state-of-the-art methods in generating vivid yet diverse unbounded 3D worlds.

Authors: Zhaoxi Chen, Guangcong Wang, Ziwei Liu.

2023-02-02

Certification of entangled quantum states and quantum measurements in Hilbert spaces of arbitrary dimension

  • The basic idea behind it is to treat a given device as a black box that given some input generates an output, and then to verify whether it works as expected by only studying the statistics generated by this device
  • The resource required in most of these certification schemes is quantum non-locality.
The emergence of quantum theory at the beginning of 20$-th$ century has changed our view of the microscopic world and has led to applications such as quantum teleportation, quantum random number generation and quantum computation to name a few, that could never have been realised using classical systems. One such application that has attracted considerable attention lately is device-independent (DI) certification of composite quantum systems. The basic idea behind it is to treat a given device as a black box that given some input generates an output, and then to verify whether it works as expected by only studying the statistics generated by this device. The novelty of these certification schemes lies in the fact that one can almost completely characterise the device (up to certain equivalences) under minimal physically well-motivated assumptions such as that the device is described using quantum theory. The resource required in most of these certification schemes is quantum non-locality. In this thesis, we construct schemes to device-independently certify quantum states and quantum measurements in Hilbert spaces of arbitrary dimension along with the optimal amount randomness that one can extract from any quantum system of arbitrary dimension.

Authors: Shubhayan Sarkar.

2023-02-02

Randomized Greedy Learning for Non-monotone Stochastic Submodular Maximization Under Full-bandit Feedback

  • We propose Randomized Greedy Learning (RGL) algorithm and theoretically prove that it achieves a $\frac{1}{2}$-regret upper bound of $\tilde{\mathcal{O}}(n T^{\frac{2}{3}})$ for horizon $T$ and number of arms $n$
  • We also show in experiments that RGL empirically outperforms other full-bandit variants in submodular and non-submodular settings.

We investigate the problem of unconstrained combinatorial multi-armed bandits with full-bandit feedback and stochastic rewards for submodular maximization. Previous works investigate the same problem assuming a submodular and monotone reward function. In this work, we study a more general problem, i.e., when the reward function is not necessarily monotone, and the submodularity is assumed only in expectation. We propose Randomized Greedy Learning (RGL) algorithm and theoretically prove that it achieves a $\frac{1}{2}$-regret upper bound of $\tilde{\mathcal{O}}(n T^{\frac{2}{3}})$ for horizon $T$ and number of arms $n$. We also show in experiments that RGL empirically outperforms other full-bandit variants in submodular and non-submodular settings.

Authors: Fares Fourati, Vaneet Aggarwal, Christopher John Quinn, Mohamed-Slim Alouini.

2023-02-02

The impact of frame quantization on the dynamic range of a one-bit image sensor

  • The discrete values of the binary rate set a maximum measurable intensity.
For a one-bit image sensor, the number of frames captured determines the quantization step size of the binary rate measurement. The discrete values of the binary rate set a maximum measurable intensity. In this note, we consider how this frame quantization impacts the high-end of the dynamic range of a one-bit image sensor.

Authors: Lucas J. Koerner.

2023-02-02

Universality in the tripartite information after global quenches: (generalised) quantum XY models

  • We map the calculation into a Riemann-Hilbert problem with a piecewise constant matrix for a doubly connected domain
  • We find an explicit solution for $\alpha=2$ and an implicit one for $\alpha>2$.

We consider the R\'enyi-$\alpha$ tripartite information $I_3^{(\alpha)}$ of three adjacent subsystems in the stationary state emerging after global quenches in noninteracting spin chains from both homogeneous and bipartite states. We identify settings in which $I_3^{(\alpha)}$ remains nonzero also in the limit of infinite lengths and develop a field theory description. We map the calculation into a Riemann-Hilbert problem with a piecewise constant matrix for a doubly connected domain. We find an explicit solution for $\alpha=2$ and an implicit one for $\alpha>2$. In the latter case, we develop a rapidly convergent perturbation theory that we use to derive analytic formulae approximating $I_3^{(\alpha)}$ with outstanding accuracy.

Authors: Vanja Marić, Maurizio Fagotti.

2023-02-02

Signatures for strong-field QED physics in the quantum limit of beamstrahlung

  • The collective radiation spectrum from a leptonic collision is derived
  • Ultrashort, ultrabright, and high-luminosity colliding gamma-ray beams are generated
  • The theoretical results are confirmed by self-consistent 3-dimensional QED particle-in-cell simulations.
Collisions between round ultrarelativistic leptonic beams are proposed to probe strong-field quantum electrodynamics (SF-QED) in the quantum limit of beamstrahlung. The collective radiation spectrum from a leptonic collision is derived. A characteristic spectral peak close to the beam energy is identified, as a signature predicted by SF-QED theory. The dependences of the spectral peak and beamstrahlung on the collision parameters are determined, paving the way to experimentally verify SF-QED. Ultrashort, ultrabright, and high-luminosity colliding gamma-ray beams are generated. The theoretical results are confirmed by self-consistent 3-dimensional QED particle-in-cell simulations.

Authors: W. L. Zhang, T. Grismayer, L. O. Silva.

2023-02-02

Generic uniqueness for the Plateau problem

  • Given a complete Riemannian manifold $\mathcal{M}\subset\mathbb{R}^d$ which is a Lipschitz neighbourhood retract of dimension $m+n$, of class $C^{3,\beta}$, without boundary and an oriented, closed submanifold $\Gamma \subset \mathcal M$ of dimension $m-1$, of class $C^{3,\alpha}$ with $\alpha<\beta$, which is a boundary in integral homology, we construct a complete metric space $\mathcal{B}$ of $C^{3,\alpha}$-perturbations of $\Gamma$ inside $\mathcal{M}$ with the following property
  • We deduce that the typical element $b\in\mathcal{B}$ admits a unique solution to the Plateau problem.

Given a complete Riemannian manifold $\mathcal{M}\subset\mathbb{R}^d$ which is a Lipschitz neighbourhood retract of dimension $m+n$, of class $C^{3,\beta}$, without boundary and an oriented, closed submanifold $\Gamma \subset \mathcal M$ of dimension $m-1$, of class $C^{3,\alpha}$ with $\alpha<\beta$, which is a boundary in integral homology, we construct a complete metric space $\mathcal{B}$ of $C^{3,\alpha}$-perturbations of $\Gamma$ inside $\mathcal{M}$ with the following property. For the typical element $b\in\mathcal B$, in the sense of Baire categories, every $m$-dimensional integral current in $\mathcal{M}$ which solves the corresponding Plateau problem has an open dense set of boundary points with density $1/2$. We deduce that the typical element $b\in\mathcal{B}$ admits a unique solution to the Plateau problem. Moreover we prove that, in a complete metric space of integral currents without boundary in $\mathbb{R}^{m+n}$, metrized by the flat norm, the typical boundary admits a unique solution to the Plateau problem.

Authors: Gianmarco Caldini, Andrea Marchese, Andrea Merlo, Simone Steinbrüchel.

2023-02-02

A classification of the Wadge hierarchies on zero-dimensional Polish spaces

  • We provide a complete classification, up to order-isomorphism, of all possible Wadge hierarchies on zero-dimensional Polish spaces using (essentially) countable ordinals as complete invariants
  • All results are based on a complete and explicit description of the Wadge hierarchy on an arbitrary zero-dimensional Polish space, depending on its topological properties.
We provide a complete classification, up to order-isomorphism, of all possible Wadge hierarchies on zero-dimensional Polish spaces using (essentially) countable ordinals as complete invariants. We also observe that although our assignment of invariants is very simple and there are only $\aleph_1$-many equivalence classes, the above classification problem is quite complex from the descriptive set-theoretic point of view: in particular, there is no Borel procedure to determine whether two zero-dimensional Polish spaces have isomorphic Wadge hierarchies. All results are based on a complete and explicit description of the Wadge hierarchy on an arbitrary zero-dimensional Polish space, depending on its topological properties.

Authors: Raphaël Carroy, Luca Motto Ros, Salvatore Scamperti.

2023-02-02

Accelerating Large Language Model Decoding with Speculative Sampling

  • We present speculative sampling, an algorithm for accelerating transformer decoding by enabling the generation of multiple tokens from each transformer call
  • This is combined with a novel modified rejection sampling scheme which preserves the distribution of the target model within hardware numerics.

We present speculative sampling, an algorithm for accelerating transformer decoding by enabling the generation of multiple tokens from each transformer call. Our algorithm relies on the observation that the latency of parallel scoring of short continuations, generated by a faster but less powerful draft model, is comparable to that of sampling a single token from the larger target model. This is combined with a novel modified rejection sampling scheme which preserves the distribution of the target model within hardware numerics. We benchmark speculative sampling with Chinchilla, a 70 billion parameter language model, achieving a 2-2.5x decoding speedup in a distributed setup, without compromising the sample quality or making modifications to the model itself.

Authors: Charlie Chen, Sebastian Borgeaud, Geoffrey Irving, Jean-Baptiste Lespiau, Laurent Sifre, John Jumper.

2023-02-02

IRIS-HEP Strategic Plan for the Next Phase of Software Upgrades for HL-LHC Physics

  • The quest to understand the fundamental building blocks of nature and their interactions is one of the oldest and most ambitious of human scientific endeavors
  • CERN's Large Hadron Collider (LHC) represents a huge step forward in this quest
  • The primary science goal is to search for physics beyond the SM and, should it be discovered, to study its implications
  • Both NSF and DOE are making large detector upgrade investments so the HL-LHC can operate in this high-rate environment.
The quest to understand the fundamental building blocks of nature and their interactions is one of the oldest and most ambitious of human scientific endeavors. CERN's Large Hadron Collider (LHC) represents a huge step forward in this quest. The discovery of the Higgs boson, the observation of exceedingly rare decays of $B$ mesons, and stringent constraints on many viable theories of physics beyond the Standard Model (SM) demonstrate the great scientific value of the LHC physics program. The next phase of this global scientific project will be the High-Luminosity LHC (HL-LHC) which will collect data starting circa 2029 and continue through the 2030s. The primary science goal is to search for physics beyond the SM and, should it be discovered, to study its implications. In the HL-LHC era, the ATLAS and CMS experiments will record around 100 times as many collisions as were used to discover the Higgs boson (and at twice the energy). Both NSF and DOE are making large detector upgrade investments so the HL-LHC can operate in this high-rate environment. Similar investment in software R&D for acquiring, managing, processing and analyzing HL-LHC data is critical to maximize the return-on-investment in the upgraded accelerator and detectors. This report presents a strategic plan for a possible second 5-year funded phase (2023 through 2028) for the Institute for Research and Innovation in Software for High Energy Physics (IRIS-HEP) which will close remaining software and computing gaps to deliver HL-LHC science.

Authors: Brian Bockelman, Peter Elmer, Gordon Watts.

2023-02-02

Are Diffusion Models Vulnerable to Membership Inference Attacks?

  • We consider both the standard diffusion models, e.g., DDPM, and the text-to-image diffusion models, e.g., Stable Diffusion
  • Experimental results demonstrate that our methods precisely infer the membership with high confidence on both of the two scenarios across six different datasets

Diffusion-based generative models have shown great potential for image synthesis, but there is a lack of research on the security and privacy risks they may pose. In this paper, we investigate the vulnerability of diffusion models to Membership Inference Attacks (MIAs), a common privacy concern. Our results indicate that existing MIAs designed for GANs or VAE are largely ineffective on diffusion models, either due to inapplicable scenarios (e.g., requiring the discriminator of GANs) or inappropriate assumptions (e.g., closer distances between synthetic images and member images). To address this gap, we propose Step-wise Error Comparing Membership Inference (SecMI), a black-box MIA that infers memberships by assessing the matching of forward process posterior estimation at each timestep. SecMI follows the common overfitting assumption in MIA where member samples normally have smaller estimation errors, compared with hold-out samples. We consider both the standard diffusion models, e.g., DDPM, and the text-to-image diffusion models, e.g., Stable Diffusion. Experimental results demonstrate that our methods precisely infer the membership with high confidence on both of the two scenarios across six different datasets

Authors: Jinhao Duan, Fei Kong, Shiqi Wang, Xiaoshuang Shi, Kaidi Xu.