Science – Page 142

Forging Faces: A New Look at Synthetic Data for Recognition

08.12.2025 by qfx

The HyperFace pipeline conjures novel identities from synthetic faces by first establishing initial embedding points-derived from a pre-trained facial recognition model-and then optimizing these embeddings on a hypersphere to maximize distinction between classes while maintaining manifold consistency, ultimately conditioning a face generator to realize the final, bespoke visages.

As demand for robust facial recognition systems grows, researchers are increasingly turning to artificially generated datasets to overcome limitations in real-world training data.

Mapping the Birth of Worlds with AI

08.12.2025 by qfx

Researchers are leveraging the power of artificial intelligence to reverse-engineer the complex processes that lead to the formation of exoplanets.

Beyond Perfect Data: Boosting Time Series Forecasts with Adversarial Training

08.12.2025 by qfx

The architecture of IdealTSF anticipates inevitable systemic failure by embracing a distributed, interconnected design, where each component’s limitations become the seeds of adaptation for the whole-a prophecy encoded in its very structure.

New research demonstrates that intentionally introducing ‘imperfect’ data can significantly improve the robustness and accuracy of time series forecasting models.

Predicting Order Book Dynamics with KANFormer

08.12.2025 by qfx

$The KANFormer architecture processes local observation blocks-snapshots $𝐗_{LOB} \in \mathbb{R}^{L\times(4n+4)}$-through dilated causal convolutions and KAN-Transformers to generate embeddings, concurrently encoding agent actions $𝐀_{actions} \in \mathbb{R}^{L\times(d+5)}$ via a KAN-Transformer, then concatenating these representations with queue position to predict density and survival functions.$

A new deep learning model leverages survival analysis to forecast the likelihood of limit orders being filled, offering a powerful tool for algorithmic trading.

Timing the Bitcoin Market: A New Model for Miner Profitability

08.12.2025 by qfx

The MineROI-Net architecture provides an overview of a system designed to efficiently process and understand mined data.

A novel deep learning framework accurately forecasts the return on investment for Bitcoin mining hardware, helping operators make smarter, data-driven decisions.

Pricing Derivatives with Smarter Labels

08.12.2025 by qfx

Predicted option prices, determined through standard, pathwise differential, and LRM differential training-each assessed with a dataset of 512 labels-demonstrate the efficacy of these training methods in approximating target prices.

A novel machine learning approach leveraging likelihood ratios significantly improves the accuracy of financial derivative pricing, particularly for complex contracts.

Modeling Complexity: When Agents Meet Neural Networks

08.12.2025 by qfx

$The agent-based model-informed neural network successfully captures the trajectories of the susceptible-infected-recovered (SIR) dynamics-defined by parameters $n=100$, $p=0.05$, $\beta=0.4$, $\gamma=0.2$, and a time step of $dt=0.1$-while simultaneously preserving the integrity of the compartmental structure inherent to the system.$

A new framework bridges the gap between agent-based simulations and deep learning to create more interpretable and reliable models of complex systems.

Can AI Beat the Market?

08.12.2025 by qfx

Large language models-specifically Zypher-7B and Mistral-7B-demonstrated the capacity to optimize mutual fund portfolios, yielding improved returns and risk-adjusted returns-particularly in Funds A and C-while Microsoft Phi-2 consistently underperformed across all tested allocations.

New research shows that artificial intelligence can intelligently allocate funds within mutual fund portfolios to maximize returns while minimizing risk.

Beyond Algorithms: Brain-Inspired Networks Predict Market Spikes

08.12.2025 by qfx

Unobserved component analysis reveals how discerning trend from differenced price series illuminates the inherent instability of economic forecasting, acknowledging that any perceived pattern is merely a temporary reprieve from inevitable divergence-a phenomenon formalized by methods like those detailed in Durbin’s work on time series and Seabold’s statistical modeling.

New research demonstrates that spiking neural networks, modeled on the human brain, can anticipate rapid price changes in high-frequency trading with improved accuracy.

Bringing Video Motion to Life: A Refinement Technique for Realistic and Diverse Movement

08.12.2025 by qfx

The proposed pipeline achieves real-time video motion transfer with reduced bandwidth requirements, enabling efficient transmission of dynamic visual information.

Researchers have developed a novel approach to enhance the realism and variability of predicted human movements in videos, achieving compelling results without the need for extensive retraining.