arXiv research

SFM resolves small-scale physics challenges in weather data.

Enhances graph classification models on small datasets.

CST-YOLO improves blood cell detection with YOLOv7 and CNN-Swin Transformer.

Study how large-scale flows align small-scale vortices in 3D Euler equations.

Study of splitting maps in Type I Ricci flows for understanding singular set structure.

The goal of this article is to draw new applications of small scale quantum ergodicity in nodal sets of eigenfunctions. We show that if quantum ergodicity holds on balls of shrinking radius $r(λ) \to 0$, then one can achieve improvements on the recent upper bounds of Logunov and Logunov-Malinnikova on the size of nodal…

PePR scores assess DL model performance per resource unit, promoting smaller, more efficient models.

In 1960 Reifenberg proved the topological disc property. He showed that a subset of $R^n$ which is well approximated by $m$-dimensional affine spaces at each point and at each (small) scale is locally a bi-Hölder image of the unit ball in $R^m$. In this paper we prove that a subset of $R^3$ which is well approximated b…

Employing profits data of Japanese companies in 2002 and 2003, we confirm that Pareto's law and the Pareto index are derived from the law of detailed balance and Gibrat's law. The last two laws are observed beyond the region where Pareto's law holds. By classifying companies into job categories, we find that companies …

Neural net reconstructs dark matter density from halo velocities.

Study on scalar curvature bounds and manifold topological complexity.

Transfer learning improves chaotic dynamics predictions with less data.

SpaPool combines dense and sparse techniques for efficient graph pooling.

We study the problem of instance segmentation in biological images with crowded and compact cells. We formulate this task as an integer program where variables correspond to cells and constraints enforce that cells do not overlap. To solve this integer program, we propose a column generation formulation where the prici…

New method improves model robustness against adversarial attacks.

A topology on a set $X$ is the same as a projection (i.e. an idempotent linear operator) $cl:2^X\to 2^X$ satisfying $A\subset cl(A)$ for all $A\subset X$. That's a good way to summarize Kuratowski's closure operator. Basic geometry on a set $X$ is a dot product $\cdot:2^X\times 2^X\to 2^Y$. Its equivalent form is an or…

This work explores how neural architecture search can improve adversarial robustness without adversarial training.

In this paper, we propose a game theoretical adversarial intervention detection mechanism for reliable smart road signs. A future trend in intelligent transportation systems is ``smart road signs" that incorporate smart codes (e.g., visible at infrared) on their surface to provide more detailed information to smart veh…

In this short note we show that the lower bounds of Mangoubi on the inner radius of nodal domains can be improved for quantum ergodic sequences of eigenfunctions, according to a certain power of the radius of shrinking balls on which the eigenfunctions equidistribute. We prove such improvements using a quick applicatio…

We study the classification of ultrametric spaces based on their small scale geometry (uniform homeomorphism), large scale geometry (coarse equivalence) and both (all scale uniform equivalences). We prove that these equivalences can be characterized with parallel constructions using a combinatoric tool called common zi…

Meta-learn Bayesian inference for task-specific BNNs using amortised inference.

In this paper we investigate the performance of different types of rectified activation functions in convolutional neural network: standard rectified linear unit (ReLU), leaky rectified linear unit (Leaky ReLU), parametric rectified linear unit (PReLU) and a new randomized leaky rectified linear units (RReLU). We evalu…

New parameterization tackles stochasticity in weather models.

Study tests rough fractional volatility model across different time scales, revealing new volatility patterns.

Fidel-TS creates a new benchmark for time series forecasting models.

Shifts dataset evaluates uncertainty in real-world tasks across modalities.

This paper is devoted to dualization of dimension-theoretical results from the small scale to the large scale. So far there are two approaches for such dualization: one consisting of creating analogs of small scale concepts and the other amounting to the covering dimension of the Higson corona $ν(X)$ of $X$. The first …

Stochastic Sparse Subspace Clustering improves subspace clustering by reducing over-segmentation through dropout.

DD-SP uses ML to improve SP for Lorenz 96 systems, outperforming LR and DD-P.

Paper proposes data quality measures for large-scale high-dimensional data.

Proposes a method for differentially private linear regression and synthetic data generation.

Efficiently applies NTK to large-scale datasets using random features.

Investigates offline RL in factorisable action spaces, overcoming overestimation bias.

Advanced Dropout improves DNN performance without requiring model-specific dropout techniques.

Deep reinforcement learning algorithms have recently been used to train multiple interacting agents in a centralised manner whilst keeping their execution decentralised. When the agents can only acquire partial observations and are faced with tasks requiring coordination and synchronisation skills, inter-agent communic…

We obtain an estimate from below for the remainder in Weyl's law on negatively curved surfaces. In the constant curvature case, such a bound was proved independently by Hejhal and Randol in 1976 using the Selberg zeta function techniques. Our approach works in arbitrary negative curvature, and is based on wave trace as…

In this study we examine the evolution of price, volume, and the bid-ask spread after extreme 15 minute intraday price changes on the NYSE and the NASDAQ. We find that due to strong behavioral trading there is an overreaction. Furthermore we find that volatility which increases sharply at the event decays according to …

Summarizing large-scaled directed graphs into small-scale representations is a useful but less studied problem setting. Conventional clustering approaches, which based on "Min-Cut"-style criteria, compress both the vertices and edges of the graph into the communities, that lead to a loss of directed edge information. O…

Reducing traffic accidents is an important public safety challenge, therefore, accident analysis and prediction has been a topic of much research over the past few decades. Using small-scale datasets with limited coverage, being dependent on extensive set of data, and being not applicable for real-time purposes are the…

This work introduces a protocol to automatically select the correct range of scales for meaningful Intrinsic Dimension estimation.

Generative Adversarial Networks (GANs) are an elegant mechanism for data generation. However, a key challenge when using GANs is how to best measure their ability to generate realistic data. In this paper, we demonstrate that an intrinsic dimensional characterization of the data space learned by a GAN model leads to an…

Deep linear networks minimize sharpness, avoiding large eigenvalues.

Constructs surfaces with specific topologies and curvatures.

Deep reinforcement learning (deep RL) has been successful in learning sophisticated behaviors automatically; however, the learning process requires a huge number of trials. In contrast, animals can learn new tasks in just a few trials, benefiting from their prior knowledge about the world. This paper seeks to bridge th…

Neuro-inspired recurrent neural network algorithms, such as echo state networks, are computationally lightweight and thereby map well onto untethered devices. The baseline echo state network algorithms are shown to be efficient in solving small-scale spatio-temporal problems. However, they underperform for complex task…

We apply state-of-the-art tools in modern high-dimensional numerical linear algebra to approximate efficiently the spectrum of the Hessian of modern deepnets, with tens of millions of parameters, trained on real data. Our results corroborate previous findings, based on small-scale networks, that the Hessian exhibits "s…

We focus in this work on the estimation of the first $k$ eigenvectors of any graph Laplacian using filtering of Gaussian random signals. We prove that we only need $k$ such signals to be able to exactly recover as many of the smallest eigenvectors, regardless of the number of nodes in the graph. In addition, we address…

Generalizes bits back coding for time-series models with latent Markov structures.