New Deep Learning Model Tackles Hierarchical Data Challenges

Researchers have developed a novel neural network architecture called Deep Generalised Mixed Models (DGMM) to address challenges in analyzing hierarchical data, particularly from experience sampling methods (ESM). This new model generalizes mixed-effects models to deep learning, allowing for flexible modeling of data structures and accommodating longitudinal outcomes with generic distributions. The DGMM utilizes an adaptation of variational auto-encoders and a Bayesian data augmentation algorithm for estimation, enabling it to scale to high-dimensional settings and provide valid inference even when data are missing at random, a common issue in studies like the GrowIt! app which tracked adolescent emotions during the COVID-19 pandemic. AI

IMPACT This new model architecture could improve the analysis of complex, longitudinal datasets, particularly in fields like psychology and social science where missing data is common.

RANK_REASON The item describes a novel neural network architecture presented in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• alphaXiv
• arXiv
• Bayesian data augmentation algorithm
• CatalyzeX Code Finder for Papers
• CORE Recommender
• COVID-19 pandemic
• DagsHub
• Deep Generalised Mixed Models
• Gotit.pub
• GrowIt!
• Hugging Face
• ScienceCast
• Variational auto-encoders with Student’s t-prior

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →