Stochastic Activation Functions Boost Financial Forecasting
A new activation function, Brownian ReLU, leverages the principles of Brownian motion to improve the performance of long short-term memory networks.
Science
Taming the Noise: A New Approach to High-Frequency Data Analysis
A robust subsampling method offers improved accuracy when estimating volatility from noisy, rapidly sampled financial data.
The Power of Anonymity: Pricing in Crowded Markets
New research shows that concealing price information can surprisingly optimize revenue in large-scale auctions with numerous participants.
Challenging the Algorithm: Humans and AI Team Up to Stress-Test Problem Solvers
![Co-FunSearch proposes a framework wherein architectural choices are not declarations of intent, but rather prophecies of eventual systemic failure, necessitating a continuous, co-evolutionary search for robust configurations rather than striving for a predetermined optimum [latex] \rightarrow \in fty [/latex].](https://arxiv.org/html/2601.16849v1/x1.png)
A novel collaboration between human intuition and artificial intelligence is proving surprisingly effective at identifying the most difficult cases for optimization algorithms.
Beyond Pixels: Super-Resolving Hyperspectral Imagery Without Real-World Data
Researchers have developed a novel unsupervised method to enhance the resolution of hyperspectral images, relying entirely on synthetic data for training.
Beyond KL: A New Approach to Spotting Anomalous Images
![The framework implements a total variation neural estimator [latex]g(x)[/latex], integrated via a dense layer [latex]g_c(h)[/latex], to refine a pre-trained classifier [latex]f[/latex], ultimately deriving both the objective function [latex]L[/latex] and an out-of-distribution (OOD) score [latex]S_c[/latex] as defined in equation 9, thereby enabling a nuanced assessment of model confidence during both training and inference.](https://arxiv.org/html/2601.15867v1/sections/TeamsIMG2002330220_1.jpeg)
Researchers are moving beyond traditional divergence measures to build more robust systems for identifying images that don’t fit the mold.
Subtle Sabotage: Injecting Hidden Attacks into Graph Networks
Researchers demonstrate a new method for launching highly effective backdoor attacks on graph classification models by subtly manipulating underlying graph structures.
The Ecosystem Play: Why Dominant Platforms Keep Prices Low
New research reveals how leading platform companies prioritize overall ecosystem strength over short-term profits in individual markets.
Synthetic Skies: Distilling Remote Sensing Data with AI
Researchers are leveraging diffusion models to create realistic, privacy-preserving synthetic datasets for remote sensing applications, offering a powerful alternative to costly and sensitive real-world data.
Turning Down the Volume on Hate: AI-Powered Music Transformation
New research explores how generative AI can subtly alter music to reduce aggressive sentiment and improve overall audio quality.