Science – Page 16

Taming Turbulence with AI: Discovering Hidden Fluid Dynamics

27.02.2026 by qfx

$Synthetic orbital trajectories, when guided by parameters reflecting periodic symmetry-such as [latex]\mathcal{T}^{s}\mathcal{R}^{a}\mathcal{S}^{m}[/latex] with varying <i>m</i> and <i>a</i>-converge to solutions of the Navier-Stokes equations, demonstrating a pathway to approximate fluid dynamics where dissipation and production rates align between synthetic (black) and converged (coloured) paths, as evidenced by trajectories reaching [latex]T\approx 1.56[/latex] and [latex]T\approx 2.46[/latex].$

Researchers are harnessing the power of deep learning to uncover previously unknown periodic orbits within turbulent flows, offering new insights into complex fluid behavior.

Sifting Signal from Noise in Single-Cell Data

27.02.2026 by qfx

The bcNMF framework facilitates a joint factorization of target and background datasets into shared, non-negative topics, enabling each sample to be represented through its topic activations and revealing underlying commonalities between the datasets.

A new approach effectively isolates key biological variations in high-dimensional datasets by actively suppressing confounding background signals.

Smarter Bidding with AI: A New Approach to Automated Advertising

27.02.2026 by qfx

$The System for Evolving Goal-Based behaviors (SEGB) operates through a three-stage process-planning a high-fidelity future state prediction [latex]s^{\prime}\_{t+1}[/latex], generating an action [latex]a^{\prime}\_{t}[/latex] conditioned on that prediction, and then refining the decision-making through offline evolution guided by a frozen critic and reference model-with online inference streamlined to the efficient planning and action generation stages.$

Researchers have developed a novel framework that leverages diffusion models and reinforcement learning to create more effective and robust automated bidding strategies for online advertising.

Beyond the Gaussian Dream: How Neural Networks Learn Features

27.02.2026 by qfx

The study demonstrates that a large-deviation rate function-derived through a posterior lens-aligns with the quadratic posterior rate induced by Gaussian-process regression utilizing the Neural Network Gaussian Process (NNGP) kernel, with the relative operator-norm gap between the selected kernel and the NNGP kernel serving as a key metric for assessing this correspondence.

New research reveals that Bayesian neural networks exhibit complex feature learning beyond simple Gaussian process behavior, offering a deeper understanding of their predictive power.

Mapping Cosmic Rays with AI: A New Direction for Radio Astronomy

27.02.2026 by qfx

The performance of the pGNN ensemble reveals a systematic bias in energy estimation at high zenith angles and correspondingly large distances of the emission point from the array, indicating a limitation in reconstructing the origin of high-energy events despite achieving a 68% energy resolution envelope around the bin-wise mean.

Researchers are leveraging the power of deep learning to pinpoint the source and energy of ultra-high-energy cosmic rays detected by ground-based radio antennas.

Beyond Tree-Boosters: Closing the Performance Gap in Tabular Data

27.02.2026 by qfx

The system explores data preprocessing strategies, alternating between Input Feature Concatenation (ICF) and Learned Feature Filtering (LFF) in each run, with ICF employing zero-padding to standardize feature vectors before concatenation-a technique designed to accommodate varying numbers of bins across different input features.

New research reveals why deep learning often lags behind tree-based methods on structured data, and introduces feature engineering techniques to level the playing field.

Decoding Machine Health: A New Path to Predictive Maintenance

27.02.2026 by qfx

The HUMS2023 dataset demonstrates earlier fault detection onset compared to prior methods.

Researchers are leveraging the power of reinforcement learning to detect machinery faults by learning what ‘normal’ operation looks like, rather than relying on scarce labeled fault data.

Beyond the Hype: AI’s Real Impact on Local News

27.02.2026 by qfx

A new study examines how artificial intelligence is being adopted by local journalism organizations and reveals a gap between expectations and practical implementation.

Building Better Trading Teams with AI

27.02.2026 by qfx

A multi-agent large language model trading system operates on the principle that all constructed systems inevitably degrade, and its architecture is designed to navigate this decay through distributed intelligence, as detailed in the accompanying text.

New research explores how carefully crafted instructions can unlock the potential of artificial intelligence to improve financial trading strategies.

Beyond Strassen: Neural Networks Learn Faster Matrix Multiplication

27.02.2026 by qfx

The experiments demonstrate that training and validation losses for matrices with entries constrained between -1 and 1 exhibit a rank-dependent behavior, with losses generally decreasing as the rank-ranging from 19 to 23-increases, as indicated by the plotted mean and standard deviation for each rank value.

Researchers are leveraging neural networks to rediscover and potentially improve upon classic algorithms for multiplying matrices, pushing the boundaries of computational efficiency.