Science – Page 21

Predicting Power Prices Through Storms and Calms

11.11.2025 by qfx

The forecast, when subjected to anomaly detection via ASM, reveals inherent instabilities within the ETT-H-1 dataset, suggesting the predictive model struggles to reconcile expected values with emergent, unpredictable fluctuations.

A new deep learning model improves the accuracy of day-ahead electricity price forecasting, even during extreme market conditions.

Designing Pensions with AI: A New Approach to Retirement Security

11.11.2025 by qfx

In a decumulation-only scenario, a neural network approach to retirement planning demonstrates markedly different outcomes compared to a provably convergent method, highlighting the potential for significant divergence in long-term financial security.

This research demonstrates how machine learning can personalize pension products by learning optimal investment strategies for diverse risk preferences.

Smarter Prompts, Better Answers: Boosting Financial AI with Synthetic Data

11.11.2025 by qfx

A system designed to refine financial prompts operates on a principle of iterative challenge, continuously generating synthetic questions, verifying their robustness, and systematically revising the prompting mechanism until error is extinguished, at which point the difficulty escalates, ensuring a perpetual cycle of refinement rather than stagnation.

A novel approach uses automatically generated data to dramatically improve the accuracy of artificial intelligence systems tackling complex financial questions.

Beyond Data Scarcity: Augmenting Molecules for Smarter Property Prediction

11.11.2025 by qfx

A new method leverages spectral graph theory to generate synthetic molecules, improving the accuracy of machine learning models for imbalanced molecular property regression.

Shaping Attacks: How Geometry Reveals Neural Network Weaknesses

11.11.2025 by qfx

A new method uses the language of shapes to expose vulnerabilities and interpret the inner workings of deep learning models.

From Field to Report: AI Automates Rocky Terrain Assessments

10.11.2025 by qfx

The automated generation of geotechnical reports proceeds not as construction, but as a cultivated flow, a methodology designed to anticipate and accommodate inevitable systemic drift within the report itself.

A new system leverages artificial intelligence to rapidly generate geotechnical reports from site data and imagery.

Hear This: Neural Networks Sharpen Audio Fingerprinting

10.11.2025 by qfx

A robust audio fingerprinting system leverages contrastive learning, where both original and intentionally degraded audio signals are processed through a shared encoder and projected into an embedding space, ultimately optimizing for invariance to common distortions like noise and reverberation.

Pretrained music models dramatically improve the accuracy and resilience of systems that identify audio, even with modifications.

Beyond the Hype: Forecasting with Graphs and Foundation Models

10.11.2025 by qfx

Time Series Forecasting Models (TSFM) operate on a five-minute sampling interval, establishing a granular temporal resolution for predictive analysis.

New research reveals that while foundation models excel at short-term forecasting, success isn’t guaranteed and depends heavily on the data itself.

Searching for Answers: The Challenge of Financial Information Retrieval

10.11.2025 by qfx

A query generation pipeline leverages both single- and multi-document approaches—the latter employing topic-based merging, context deepening, and comparative analysis—to produce inquiries subsequently vetted for quality and demonstrable answerability.

New research reveals current systems struggle with complex questions requiring synthesis from multiple financial documents.

Spotting the Flaws in Fake Data: A New Approach to Medical Image Quality

10.11.2025 by qfx

The proposed method rigorously identifies distortions in data attributable to network effects, thereby enabling precise artifact detection.

Researchers develop a method to automatically identify and remove artificial distortions in synthetic medical images, boosting their usefulness for training AI.