Search

Splatting-based 3D reconstruction methods have gained popularity with the advent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel views. These methods commonly resort to using exponential family functions, such as the Gaussian function, as reconstruction kernels due to their anisotropic nature, ease of projection, and differentiability in rasterization. However, the field remains restricted to variations within the exponential family, leaving generalized reconstruction kernels largely underexplored, partly due to the lack of easy integrability in 3D to 2D projections. In this light, we show that a class of decaying anisotropic radial basis functions (DARBFs), which are non-negative functions of the Mahalanobis distance, supports splatting by approximating the Gaussian function's closed-form integration advantage. With this fresh perspective, we demonstrate up to 34% faster convergence during training and a 15% reduction in memory consumption across various DARB reconstruction kernels, while maintaining comparable PSNR, SSIM, and LPIPS results. We will make the code available., Comment: Link to the project page: https://randomnerds.github.io/darbs.github.io/

We present a spatio-temporal perspective on category-agnostic 3D lifting of 2D keypoints over a temporal sequence. Our approach differs from existing state-of-the-art methods that are either: (i) object agnostic, but can only operate on individual frames, or (ii) can model space-time dependencies, but are only designed to work with a single object category. Our approach is grounded in two core principles. First, when there is a lack of data about an object, general information from similar objects can be leveraged for better performance. Second, while temporal information is important, the most critical information is in immediate temporal proximity. These two principles allow us to outperform current state-of-the-art methods on per-frame and per-sequence metrics for a variety of objects. Lastly, we release a new synthetic dataset containing 3D skeletons and motion sequences of a diverse set animals. Dataset and code will be made publicly available., Comment: 3DV 2025

This paper challenges the conventional belief that softmax attention in transformers is effective primarily because it generates a probability distribution for attention allocation. Instead, we theoretically show that its success lies in its ability to implicitly regularize the Frobenius norm of the attention matrix during training. We then explore alternative activations that regularize the Frobenius norm of the attention matrix, demonstrating that certain polynomial activations can achieve this effect, making them suitable for attention-based architectures. Empirical results indicate these activations perform comparably or better than softmax across various computer vision and language tasks, suggesting new possibilities for attention mechanisms beyond softmax.

$\Lambda$CDM cosmology predicts the hierarchical formation of galaxies which build up mass by merger events and accreting smaller systems. The stellar halo of the Milky Way has proven to be useful a tool for tracing this accretion history. However, most of this work has focused on the outer halo where dynamical times are large and the dynamical properties of accreted systems are preserved. In this work, we investigate the inner galaxy regime, where dynamical times are relatively small and systems are generally completely phase-mixed. Using the FIRE-2 and Auriga cosmological zoom-in simulation suites of Milky Way-mass galaxies, we find the stellar density profiles along the minor axis (perpendicular to the galactic disk) within the NFW scale radii (R$\approx$15 kpc) are best described as an exponential disk with scale height <0.3 kpc and a power law component with slope $\alpha\approx$-4. The stellar density amplitude and slope for the power law component is not significantly correlated with metrics of the galaxy's accretion history. Instead, we find the stellar profiles strongly correlate with the dark matter profile. Across simulation suites, the galaxies studied in this work have a stellar to dark matter mass ratio that decreases as $1/r^2$ along the minor axis., Comment: 29 pages, 13 figures, 1 table, Submitted to ApJ, comments welcome!

In this article, we introduce a novel normalization technique for neural network weight matrices, which we term weight conditioning. This approach aims to narrow the gap between the smallest and largest singular values of the weight matrices, resulting in better-conditioned matrices. The inspiration for this technique partially derives from numerical linear algebra, where well-conditioned matrices are known to facilitate stronger convergence results for iterative solvers. We provide a theoretical foundation demonstrating that our normalization technique smoothens the loss landscape, thereby enhancing convergence of stochastic gradient descent algorithms. Empirically, we validate our normalization across various neural network architectures, including Convolutional Neural Networks (CNNs), Vision Transformers (ViT), Neural Radiance Fields (NeRF), and 3D shape modeling. Our findings indicate that our normalization method is not only competitive but also outperforms existing weight normalization techniques from the literature., Comment: ECCV 2024

The Dark Energy Spectroscopic Instrument (DESI) Peculiar Velocity Survey aims to measure the peculiar velocities of early and late type galaxies within the DESI footprint using both the Fundamental Plane and Tully-Fisher relations. Direct measurements of peculiar velocities can significantly improve constraints on the growth rate of structure, reducing uncertainty by a factor of approximately 2.5 at redshift 0.1 compared to the DESI Bright Galaxy Survey's redshift space distortion measurements alone. We assess the quality of stellar velocity dispersion measurements from DESI spectroscopic data. These measurements, along with photometric data from the Legacy Survey, establish the Fundamental Plane relation and determine distances and peculiar velocities of early-type galaxies. During Survey Validation, we obtain spectra for 6698 unique early-type galaxies, up to a photometric redshift of 0.15. 64\% of observed galaxies (4267) have relative velocity dispersion errors below 10\%. This percentage increases to 75\% if we restrict our sample to galaxies with spectroscopic redshifts below 0.1. We use the measured central velocity dispersion, along with photometry from the DESI Legacy Imaging Surveys, to fit the Fundamental Plane parameters using a 3D Gaussian maximum likelihood algorithm that accounts for measurement uncertainties and selection cuts. In addition, we conduct zero-point calibration using the absolute distance measurements to the Coma cluster, leading to a value of the Hubble constant, $H_0 = 76.05 \pm 0.35$(statistical) $\pm 0.49$(systematic FP) $\pm 4.86$(statistical due to calibration) $\mathrm{km \ s^{-1} Mpc^{-1}}$. This $H_0$ value is within $2\sigma$ of Planck Cosmic Microwave Background results and within $1\sigma$, of other low redshift distance indicator-based measurements., Comment: 18 pages, 9 figures, 2 tables. Submitted for publication in MNRAS

This paper tackles the simultaneous optimization of pose and Neural Radiance Fields (NeRF). Departing from the conventional practice of using explicit global representations for camera pose, we propose a novel overparameterized representation that models camera poses as learnable rigid warp functions. We establish that modeling the rigid warps must be tightly coupled with constraints and regularization imposed. Specifically, we highlight the critical importance of enforcing invertibility when learning rigid warp functions via neural network and propose the use of an Invertible Neural Network (INN) coupled with a geometry-informed constraint for this purpose. We present results on synthetic and real-world datasets, and demonstrate that our approach outperforms existing baselines in terms of pose estimation and high-fidelity reconstruction due to enhanced optimization convergence., Comment: Accepted to ECCV 2024. Project page: https://sfchng.github.io/ineurowarping-github.io/

We present a method for dynamic surface reconstruction of large-scale urban scenes from LiDAR. Depth-based reconstructions tend to focus on small-scale objects or large-scale SLAM reconstructions that treat moving objects as outliers. We take a holistic perspective and optimize a compositional model of a dynamic scene that decomposes the world into rigidly-moving objects and the background. To achieve this, we take inspiration from recent novel view synthesis methods and frame the reconstruction problem as a global optimization over neural surfaces, ego poses, and object poses, which minimizes the error between composed spacetime surfaces and input LiDAR scans. In contrast to view synthesis methods, which typically minimize 2D errors with gradient descent, we minimize a 3D point-to-surface error by coordinate descent, which we decompose into registration and surface reconstruction steps. Each step can be handled well by off-the-shelf methods without any re-training. We analyze the surface reconstruction step for rolling-shutter LiDARs, and show that deskewing operations common in continuous time SLAM can be applied to dynamic objects as well, improving results over prior art by an order of magnitude. Beyond pursuing dynamic reconstruction as a goal in and of itself, we propose that such a system can be used to auto-label partially annotated sequences and produce ground truth annotation for hard-to-label problems such as depth completion and scene flow., Comment: To appear in 3DV 2025

Empirical correlations connecting starlight to galaxy dynamics (e.g., the fundamental plane (FP) of elliptical/quiescent galaxies and the Tully--Fisher relation of spiral/star-forming galaxies) provide cosmology-independent distance estimation and are central to local Universe cosmology. In this work, we introduce the mass hyperplane (MH), which is the stellar-to-dynamical mass relation $(M_\star/M_\mathrm{dyn})$ recast as a linear distance indicator. Building on recent FP studies, we show that both star-forming and quiescent galaxies follow the same empirical MH, then use this to measure the peculiar velocities (PVs) for a sample of 2496 galaxies at $z<0.12$ from GAMA. The limiting precision of MH-derived distance/PV estimates is set by the intrinsic scatter in size, which we find to be $\approx$0.1~dex for both quiescent and star-forming galaxies (when modeled independently) and $\approx$0.11~dex when all galaxies are modeled together; showing that the MH is as good as the FP. To empirically validate our framework and distance/PV estimates, we compare the inferred distances to groups as derived using either quiescent or star-forming galaxies. A good agreement is obtained with no discernible bias or offset, having a scatter of $\approx$0.05~dex $\approx$12\% in distance. Further, we compare our PV measurements for the quiescent galaxies to the previous PV measurements of the galaxies in common between GAMA and the Sloan Digital Sky Survey (SDSS), which shows similarly good agreement. Finally, we provide comparisons of PV measurements made with the FP and the MH, then discuss possible improvements in the context of upcoming surveys such as the 4MOST Hemisphere Survey (4HS)., Comment: Accepted: 15th May 2024

The training of vision transformer (ViT) networks on small-scale datasets poses a significant challenge. By contrast, convolutional neural networks (CNNs) have an architectural inductive bias enabling them to perform well on such problems. In this paper, we argue that the architectural bias inherent to CNNs can be reinterpreted as an initialization bias within ViT. This insight is significant as it empowers ViTs to perform equally well on small-scale problems while maintaining their flexibility for large-scale applications. Our inspiration for this ``structured'' initialization stems from our empirical observation that random impulse filters can achieve comparable performance to learned filters within CNNs. Our approach achieves state-of-the-art performance for data-efficient ViT learning across numerous benchmarks including CIFAR-10, CIFAR-100, and SVHN., Comment: 20 pages, 5 figures, 8 tables

Low-rank decomposition has emerged as a vital tool for enhancing parameter efficiency in neural network architectures, gaining traction across diverse applications in machine learning. These techniques significantly lower the number of parameters, striking a balance between compactness and performance. However, a common challenge has been the compromise between parameter efficiency and the accuracy of the model, where reduced parameters often lead to diminished accuracy compared to their full-rank counterparts. In this work, we propose a novel theoretical framework that integrates a sinusoidal function within the low-rank decomposition process. This approach not only preserves the benefits of the parameter efficiency characteristic of low-rank methods but also increases the decomposition's rank, thereby enhancing model performance. Our method proves to be a plug in enhancement for existing low-rank models, as evidenced by its successful application in Vision Transformers (ViT), Large Language Models (LLMs), Neural Radiance Fields (NeRF) and 3D shape modelling., Comment: The first two authors contributed equally. Paper accepted at ICLR 2025

In the realm of computer vision, Neural Fields have gained prominence as a contemporary tool harnessing neural networks for signal representation. Despite the remarkable progress in adapting these networks to solve a variety of problems, the field still lacks a comprehensive theoretical framework. This article aims to address this gap by delving into the intricate interplay between initialization and activation, providing a foundational basis for the robust optimization of Neural Fields. Our theoretical insights reveal a deep-seated connection among network initialization, architectural choices, and the optimization process, emphasizing the need for a holistic approach when designing cutting-edge Neural Fields., Comment: CVPR 2024

Deep implicit functions have been found to be an effective tool for efficiently encoding all manner of natural signals. Their attractiveness stems from their ability to compactly represent signals with little to no offline training data. Instead, they leverage the implicit bias of deep networks to decouple hidden redundancies within the signal. In this paper, we explore the hypothesis that additional compression can be achieved by leveraging redundancies that exist between layers. We propose to use a novel runtime decoder-only hypernetwork - that uses no offline training data - to better exploit cross-layer parameter redundancy. Previous applications of hypernetworks with deep implicit functions have employed feed-forward encoder/decoder frameworks that rely on large offline datasets that do not generalize beyond the signals they were trained on. We instead present a strategy for the optimization of runtime deep implicit functions for single-instance signals through a Decoder-Only randomly projected Hypernetwork (D'OH). By directly changing the latent code dimension, we provide a natural way to vary the memory footprint of neural representations without the costly need for neural architecture search on a space of alternative low-rate structures., Comment: 29 pages, 17 figures

In contrast to current state-of-the-art methods, such as NSFP [25], which employ deep implicit neural functions for modeling scene flow, we present a novel approach that utilizes classical kernel representations. This representation enables our approach to effectively handle dense lidar points while demonstrating exceptional computational efficiency -- compared to recent deep approaches -- achieved through the solution of a linear system. As a runtime optimization-based method, our model exhibits impressive generalizability across various out-of-distribution scenarios, achieving competitive performance on large-scale lidar datasets. We propose a new positional encoding-based kernel that demonstrates state-of-the-art performance in efficient lidar scene flow estimation on large-scale point clouds. An important highlight of our method is its near real-time performance (~150-170 ms) with dense lidar data (~8k-144k points), enabling a variety of practical applications in robotics and autonomous driving scenarios., Comment: 23 pages, 8 figures, 3 tables

We tackle semi-supervised object detection based on motion cues. Recent results suggest that heuristic-based clustering methods in conjunction with object trackers can be used to pseudo-label instances of moving objects and use these as supervisory signals to train 3D object detectors in Lidar data without manual supervision. We re-think this approach and suggest that both, object detection, as well as motion-inspired pseudo-labeling, can be tackled in a data-driven manner. We leverage recent advances in scene flow estimation to obtain point trajectories from which we extract long-term, class-agnostic motion patterns. Revisiting correlation clustering in the context of message passing networks, we learn to group those motion patterns to cluster points to object instances. By estimating the full extent of the objects, we obtain per-scan 3D bounding boxes that we use to supervise a Lidar object detection network. Our method not only outperforms prior heuristic-based approaches (57.5 AP, +14 improvement over prior work), more importantly, we show we can pseudo-label and train object detectors across datasets., Comment: Accepted to CVPR 2024!

Implicit neural representations have emerged as a powerful technique for encoding complex continuous multidimensional signals as neural networks, enabling a wide range of applications in computer vision, robotics, and geometry. While Adam is commonly used for training due to its stochastic proficiency, it entails lengthy training durations. To address this, we explore alternative optimization techniques for accelerated training without sacrificing accuracy. Traditional second-order optimizers like L-BFGS are suboptimal in stochastic settings, making them unsuitable for large-scale data sets. Instead, we propose stochastic training using curvature-aware diagonal preconditioners, showcasing their effectiveness across various signal modalities such as images, shape reconstruction, and Neural Radiance Fields (NeRF)., Comment: The first two authors contributed equally

Implicit Neural Representations (INRs) have gained popularity for encoding signals as compact, differentiable entities. While commonly using techniques like Fourier positional encodings or non-traditional activation functions (e.g., Gaussian, sinusoid, or wavelets) to capture high-frequency content, their properties lack exploration within a unified theoretical framework. Addressing this gap, we conduct a comprehensive analysis of these activations from a sampling theory perspective. Our investigation reveals that sinc activations, previously unused in conjunction with INRs, are theoretically optimal for signal encoding. Additionally, we establish a connection between dynamical systems and INRs, leveraging sampling theory to bridge these two paradigms.

Recently, neural networks utilizing periodic activation functions have been proven to demonstrate superior performance in vision tasks compared to traditional ReLU-activated networks. However, there is still a limited understanding of the underlying reasons for this improved performance. In this paper, we aim to address this gap by providing a theoretical understanding of periodically activated networks through an analysis of their Neural Tangent Kernel (NTK). We derive bounds on the minimum eigenvalue of their NTK in the finite width setting, using a fairly general network architecture which requires only one wide layer that grows at least linearly with the number of data samples. Our findings indicate that periodically activated networks are \textit{notably more well-behaved}, from the NTK perspective, than ReLU activated networks. Additionally, we give an application to the memorization capacity of such networks and verify our theoretical predictions empirically. Our study offers a deeper understanding of the properties of periodically activated neural networks and their potential in the field of deep learning., Comment: arXiv admin note: substantial text overlap with arXiv:2402.02711

Physics-informed neural networks (PINNs) offer a promising avenue for tackling both forward and inverse problems in partial differential equations (PDEs) by incorporating deep learning with fundamental physics principles. Despite their remarkable empirical success, PINNs have garnered a reputation for their notorious training challenges across a spectrum of PDEs. In this work, we delve into the intricacies of PINN optimization from a neural architecture perspective. Leveraging the Neural Tangent Kernel (NTK), our study reveals that Gaussian activations surpass several alternate activations when it comes to effectively training PINNs. Building on insights from numerical linear algebra, we introduce a preconditioned neural architecture, showcasing how such tailored architectures enhance the optimization process. Our theoretical findings are substantiated through rigorous validation against established PDEs within the scientific literature.

Training vision transformer networks on small datasets poses challenges. In contrast, convolutional neural networks (CNNs) can achieve state-of-the-art performance by leveraging their architectural inductive bias. In this paper, we investigate whether this inductive bias can be reinterpreted as an initialization bias within a vision transformer network. Our approach is motivated by the finding that random impulse filters can achieve almost comparable performance to learned filters in CNNs. We introduce a novel initialization strategy for transformer networks that can achieve comparable performance to CNNs on small datasets while preserving its architectural flexibility., Comment: 14 pages, 9 figures, 8 tables

The lifting of 3D structure and camera from 2D landmarks is at the cornerstone of the entire discipline of computer vision. Traditional methods have been confined to specific rigid objects, such as those in Perspective-n-Point (PnP) problems, but deep learning has expanded our capability to reconstruct a wide range of object classes (e.g. C3DPO and PAUL) with resilience to noise, occlusions, and perspective distortions. All these techniques, however, have been limited by the fundamental need to establish correspondences across the 3D training data -- significantly limiting their utility to applications where one has an abundance of "in-correspondence" 3D data. Our approach harnesses the inherent permutation equivariance of transformers to manage varying number of points per 3D data instance, withstands occlusions, and generalizes to unseen categories. We demonstrate state of the art performance across 2D-3D lifting task benchmarks. Since our approach can be trained across such a broad class of structures we refer to it simply as a 3D Lifting Foundation Model (3D-LFM) -- the first of its kind., Comment: Visit the project page at https://3dlfm.github.io for links to additional media, code, and videos. The site also features a custom GPT tailored to address queries related to 3D-LFM. Accepted at CVPR 2024

The test-time optimization of scene flow - using a coordinate network as a neural prior - has gained popularity due to its simplicity, lack of dataset bias, and state-of-the-art performance. We observe, however, that although coordinate networks capture general motions by implicitly regularizing the scene flow predictions to be spatially smooth, the neural prior by itself is unable to identify the underlying multi-body rigid motions present in real-world data. To address this, we show that multi-body rigidity can be achieved without the cumbersome and brittle strategy of constraining the $SE(3)$ parameters of each rigid body as done in previous works. This is achieved by regularizing the scene flow optimization to encourage isometry in flow predictions for rigid bodies. This strategy enables multi-body rigidity in scene flow while maintaining a continuous flow field, hence allowing dense long-term scene flow integration across a sequence of point clouds. We conduct extensive experiments on real-world datasets and demonstrate that our approach outperforms the state-of-the-art in 3D scene flow and long-term point-wise 4D trajectory prediction. The code is available at: https://github.com/kavisha725/MBNSF., Comment: Accepted for 3DV'2024 (oral)

End-to-end trained per-point embeddings are an essential ingredient of any state-of-the-art 3D point cloud processing such as detection or alignment. Methods like PointNet, or the more recent point cloud transformer -- and its variants -- all employ learned per-point embeddings. Despite impressive performance, such approaches are sensitive to out-of-distribution (OOD) noise and outliers. In this paper, we explore the role of an analytical per-point embedding based on the criterion of bandwidth. The concept of bandwidth enables us to draw connections with an alternate per-point embedding -- positional embedding, particularly random Fourier features. We present compelling robust results across downstream tasks such as point cloud classification and registration with several categories of OOD noise., Comment: 18 pages, 13 figures, 5 tables

We advocate for a Galactic center (GC) field to be added to the Galactic Bulge Time Domain Survey (GBTDS). The new field would yield high-cadence photometric and astrometric measurements of an unprecedented ${\sim}$3.3 million stars toward the GC. This would enable a wide range of science cases, such as finding star-compact object binaries that may ultimately merge as LISA-detectable gravitational wave sources, constraining the mass function of stars and compact objects in different environments, detecting populations of microlensing and transiting exoplanets, studying stellar flares and variability in young and old stars, and monitoring accretion onto the central supermassive black hole. In addition, high-precision proper motions and parallaxes would open a new window into the large-scale dynamics of stellar populations at the GC, yielding insights into the formation and evolution of galactic nuclei and their co-evolution with the growth of the supermassive black hole. We discuss the possible trade-offs between the notional GBTDS and the addition of a GC field with either an optimal or minimal cadence. Ultimately, the addition of a GC field to the GBTDS would dramatically increase the science return of Roman and provide a legacy dataset to study the mid-plane and innermost regions of our Galaxy., Comment: 19 pages, 3 figures. Submitted to the NASA Roman Core Community Surveys White Paper Call

We present a new approach to understanding the relationship between loss curvature and input-output model behaviour in deep learning. Specifically, we use existing empirical analyses of the spectrum of deep network loss Hessians to ground an ansatz tying together the loss Hessian and the input-output Jacobian over training samples during the training of deep neural networks. We then prove a series of theoretical results which quantify the degree to which the input-output Jacobian of a model approximates its Lipschitz norm over a data distribution, and deduce a novel generalisation bound in terms of the empirical Jacobian. We use our ansatz, together with our theoretical results, to give a new account of the recently observed progressive sharpening phenomenon, as well as the generalisation properties of flat minima. Experimental evidence is provided to validate our claims.

Coordinate networks are widely used in computer vision due to their ability to represent signals as compressed, continuous entities. However, training these networks with first-order optimizers can be slow, hindering their use in real-time applications. Recent works have opted for shallow voxel-based representations to achieve faster training, but this sacrifices memory efficiency. This work proposes a solution that leverages second-order optimization methods to significantly reduce training times for coordinate networks while maintaining their compressibility. Experiments demonstrate the effectiveness of this approach on various signal modalities, such as audio, images, videos, shape reconstruction, and neural radiance fields.

Neural Scene Flow Prior (NSFP) is of significant interest to the vision community due to its inherent robustness to out-of-distribution (OOD) effects and its ability to deal with dense lidar points. The approach utilizes a coordinate neural network to estimate scene flow at runtime, without any training. However, it is up to 100 times slower than current state-of-the-art learning methods. In other applications such as image, video, and radiance function reconstruction innovations in speeding up the runtime performance of coordinate networks have centered upon architectural changes. In this paper, we demonstrate that scene flow is different -- with the dominant computational bottleneck stemming from the loss function itself (i.e., Chamfer distance). Further, we rediscover the distance transform (DT) as an efficient, correspondence-free loss function that dramatically speeds up the runtime optimization. Our fast neural scene flow (FNSF) approach reports for the first time real-time performance comparable to learning methods, without any training or OOD bias on two of the largest open autonomous driving (AV) lidar datasets Waymo Open and Argoverse., Comment: 17 pages, 11 figures, 6 tables

Popular benchmarks for self-supervised LiDAR scene flow (stereoKITTI, and FlyingThings3D) have unrealistic rates of dynamic motion, unrealistic correspondences, and unrealistic sampling patterns. As a result, progress on these benchmarks is misleading and may cause researchers to focus on the wrong problems. We evaluate a suite of top methods on a suite of real-world datasets (Argoverse 2.0, Waymo, and NuScenes) and report several conclusions. First, we find that performance on stereoKITTI is negatively correlated with performance on real-world data. Second, we find that one of this task's key components -- removing the dominant ego-motion -- is better solved by classic ICP than any tested method. Finally, we show that despite the emphasis placed on learning, most performance gains are caused by pre- and post-processing steps: piecewise-rigid refinement and ground removal. We demonstrate this through a baseline method that combines these processing steps with a learning-free test-time flow optimization. This baseline outperforms every evaluated method., Comment: WACV 2024

In this paper we present a new method for deformable NeRF that can directly use optical flow as supervision. We overcome the major challenge with respect to the computationally inefficiency of enforcing the flow constraints to the backward deformation field, used by deformable NeRFs. Specifically, we show that inverting the backward deformation function is actually not needed for computing scene flows between frames. This insight dramatically simplifies the problem, as one is no longer constrained to deformation functions that can be analytically inverted. Instead, thanks to the weak assumptions required by our derivation based on the inverse function theorem, our approach can be extended to a broad class of commonly used backward deformation field. We present results on monocular novel view synthesis with rapid object motion, and demonstrate significant improvements over baselines without flow supervision.

Supermassive black holes (SMBHs) are a key catalyst of galaxy formation and evolution, leading to an observed correlation between SMBH mass $M_{\rm BH}$ and host galaxy velocity dispersion $\sigma_{\rm e}$. Outside the local Universe, measurements of $M_{\rm BH}$ are usually only possible for SMBHs in an active state: limiting sample size and introducing selection biases. Gravitational lensing makes it possible to measure the mass of non-active SMBHs. We present models of the $z=0.169$ galaxy-scale strong lens Abell~1201. A cD galaxy in a galaxy cluster, it has sufficient `external shear' that a magnified image of a $z = 0.451$ background galaxy is projected just $\sim 1$ kpc from the galaxy centre. Using multi-band Hubble Space Telescope imaging and the lens modeling software $\texttt{PyAutoLens}$ we reconstruct the distribution of mass along this line of sight. Bayesian model comparison favours a point mass with $M_{\rm BH} = 3.27 \pm 2.12\times10^{10}\,$M$_{\rm \odot}$ (3$\sigma$ confidence limit); an ultramassive black hole. One model gives a comparable Bayesian evidence without a SMBH, however we argue this model is nonphysical given its base assumptions. This model still provides an upper limit of $M_{\rm BH} \leq 5.3 \times 10^{10}\,$M$_{\rm \odot}$, because a SMBH above this mass deforms the lensed image $\sim 1$ kpc from Abell 1201's centre. This builds on previous work using central images to place upper limits on $M_{\rm BH}$, but is the first to also place a lower limit and without a central image being observed. The success of this method suggests that surveys during the next decade could measure thousands more SMBH masses, and any redshift evolution of the $M_{\rm BH}$--$\sigma_{\rm e}$ relation. Results are available at https://github.com/Jammy2211/autolens_abell_1201., Comment: Accepted in MNRAS, 27 pages, 22 figures

Modelling dynamical systems is an integral component for understanding the natural world. To this end, neural networks are becoming an increasingly popular candidate owing to their ability to learn complex functions from large amounts of data. Despite this recent progress, there has not been an adequate discussion on the architectural regularization that neural networks offer when learning such systems, hindering their efficient usage. In this paper, we initiate a discussion in this direction using coordinate networks as a test bed. We interpret dynamical systems and coordinate networks from a signal processing lens, and show that simple coordinate networks with few layers can be used to solve multiple problems in modelling dynamical systems, without any explicit regularizers.

We describe the target selection and characteristics of the DESI Peculiar Velocity Survey, the largest survey of peculiar velocities (PVs) using both the fundamental plane (FP) and the Tully-Fisher (TF) relationship planned to date. We detail how we identify suitable early-type galaxies (ETGs) for the FP and suitable late-type galaxies (LTGs) for the TF relation using the photometric data provided by the DESI Legacy Imaging Survey DR9. Subsequently, we provide targets for 373 533 ETGs and 118 637 LTGs within the DESI 5-year footprint. We validate these photometric selections using existing morphological classifications. Furthermore, we demonstrate using survey validation data that DESI is able to measure the spectroscopic properties to sufficient precision to obtain PVs for our targets. Based on realistic DESI fiber assignment simulations and spectroscopic success rates, we predict the final DESI PV Survey will obtain $\sim$133 000 FP-based and $\sim$53 000 TF-based PV measurements over an area of 14 000 $\mathrm{deg^{2}}$. We forecast the ability of using these data to measure the clustering of galaxy positions and PVs from the combined DESI PV and Bright Galaxy Surveys (BGS), which allows for cancellation of cosmic variance at low redshifts. With these forecasts, we anticipate a $4\%$ statistical measurement on the growth rate of structure at $z<0.15$. This is over two times better than achievable with redshifts from the BGS alone. The combined DESI PV and BGS will enable the most precise tests to date of the time and scale dependence of large-scale structure growth at $z<0.15$., Comment: 20 pages, 15 figures, 14 tables; accepted in MNRAS

The amount and distribution of water on the lunar surface are related to the input and production of water by solar wind and meteoroid bombardment, balanced by photodestruction and mobility across the surface. Using the Stratospheric Observatory for Infrared Astronomy (SOFIA), we imaged the 6.1 micron feature that uniquely traces molecular water, covering 1/4 of the lunar nearside surface south of -60 degrees latitude with 5 km resolution on 2022 Feb 17 UTC. The water feature strength varies significantly across the region, being drier at +28 degrees longitude to more wet (~170 ppm) at -7 degrees longitude, and also decreasing toward the pole. Significant local enhancements are found, associated with south-facing, high-altitude topographic features. This includes relatively high H2O concentration in a "wet ridge" just north of Curtius crater; the south-facing, northern, inner rims of most prominent craters; the south face of the central peak of Moretus crater; and permanently-shadowed polar regions., Comment: accepted to Planetary Science Journal on 2/20/2023

We present a new technique to perform passive bistatic subsurface radar probes on airless planetary bodies. This technique uses the naturally occurring radio impulses generated when high-energy cosmic rays impact the body's surface. As in traditional radar sounding, the downward-beamed radio emission from each individual cosmic ray impact will reflect off subsurface dielectric contrasts and propagate back up to the surface to be detected. We refer to this technique as Askaryan radar after the fundamental physics process, the Askaryan effect, that produces this radio emission. This technique can be performed from an orbiting satellite, or from a surface lander, but since the radio emission is generated beneath the surface, an Askaryan radar can completely bypass the effects of surface clutter and backscatter typically associated with surface-penetrating radar. We present the background theory of Askaryan subsurface radar and show results from both finite-difference time-domain (FDTD) and Monte Carlo simulations that confirm that this technique is a promising planetary radar sounding method, producing detectable signals for realistic planetary science applications., Comment: 18 pages, 5 figures

We introduce a general theoretical framework, designed for the study of gradient optimisation of deep neural networks, that encompasses ubiquitous architecture choices including batch normalisation, weight normalisation and skip connections. Our framework determines the curvature and regularity properties of multilayer loss landscapes in terms of their constituent layers, thereby elucidating the roles played by normalisation layers and skip connections in globalising these properties. We then demonstrate the utility of this framework in two respects. First, we give the only proof of which we are aware that a class of deep neural networks can be trained using gradient descent to global optima even when such optima only exist at infinity, as is the case for the cross-entropy cost. Second, we identify a novel causal mechanism by which skip connections accelerate training, which we verify predictively with ResNets on MNIST, CIFAR10, CIFAR100 and ImageNet., Comment: NeurIPS 2023

Labeling articulated objects in unconstrained settings have a wide variety of applications including entertainment, neuroscience, psychology, ethology, and many fields of medicine. Large offline labeled datasets do not exist for all but the most common articulated object categories (e.g., humans). Hand labeling these landmarks within a video sequence is a laborious task. Learned landmark detectors can help, but can be error-prone when trained from only a few examples. Multi-camera systems that train fine-grained detectors have shown significant promise in detecting such errors, allowing for self-supervised solutions that only need a small percentage of the video sequence to be hand-labeled. The approach, however, is based on calibrated cameras and rigid geometry, making it expensive, difficult to manage, and impractical in real-world scenarios. In this paper, we address these bottlenecks by combining a non-rigid 3D neural prior with deep flow to obtain high-fidelity landmark estimates from videos with only two or three uncalibrated, handheld cameras. With just a few annotations (representing 1-2% of the frames), we are able to produce 2D results comparable to state-of-the-art fully supervised methods, along with 3D reconstructions that are impossible with other existing approaches. Our Multi-view Bootstrapping in the Wild (MBW) approach demonstrates impressive results on standard human datasets, as well as tigers, cheetahs, fish, colobus monkeys, chimpanzees, and flamingos from videos captured casually in a zoo. We release the codebase for MBW as well as this challenging zoo dataset consisting image frames of tail-end distribution categories with their corresponding 2D, 3D labels generated from minimal human intervention., Comment: NeurIPS 2022 conference. Project webpage and code: https://github.com/mosamdabhi/MBW

The role of quantization within implicit/coordinate neural networks is still not fully understood. We note that using a canonical fixed quantization scheme during training produces poor performance at low-rates due to the network weight distributions changing over the course of training. In this work, we show that a non-uniform quantization of neural weights can lead to significant improvements. Specifically, we demonstrate that a clustered quantization enables improved reconstruction. Finally, by characterising a trade-off between quantization and network capacity, we demonstrate that it is possible (while memory inefficient) to reconstruct signals using binary neural networks. We demonstrate our findings experimentally on 2D image reconstruction and 3D radiance fields; and show that simple quantization methods and architecture search can achieve compression of NeRF to less than 16kb with minimal loss in performance (323x smaller than the original NeRF)., Comment: 10 pages, 10 figures