PMB-NN: Physiology-Centred Hybrid AI for Personalized Hemodynamic Monitoring from Photoplethysmography

A recent survey published on arXiv explores the concept of symmetry in neural network parameter spaces, highlighting how modern deep learning models exhibit significant overparameterization. This redundancy is largely attributed to symmetries that maintain the network's output unchanged, influencing optimization and learning dynamics.

The proposed CIEGAD framework aims to enhance data augmentation in deep learning by addressing the challenges of data scarcity and label imbalance, which often lead to misclassification and unstable model behavior. By employing cluster conditioning and hierarchical frequency allocation, CIEGAD systematically improves both in-distribution and out-of-distribution data regions.

A new framework for evaluating AI metacognition has been proposed, focusing on metacognitive sensitivity, which assesses how reliably a model's confidence predicts its accuracy. This framework introduces a dynamic sensitivity score that informs a bandit-based arbiter for test-time model selection, enhancing the decision-making process in deep learning models such as CNNs and VLMs.

Pulse3DFace has been introduced as the first dataset for estimating 3D blood pulsation maps, enabling the development of models for dynamic facial blood pulsation. This dataset includes raw videos from 15 subjects, recorded from 23 viewpoints, along with processed 3D pulsation maps compatible with the FLAME 3D head model.

A recent study has introduced a robust multi-disease retinal classification system utilizing Xception-based transfer learning and W-Net vessel segmentation, addressing the increasing incidence of vision-threatening ocular conditions. This approach combines deep feature extraction with interpretable image processing to enhance the accuracy of automated diagnoses.

A new framework called 3D Inverse Design (3DID) has been proposed to optimize aerodynamic designs directly in three-dimensional space, overcoming limitations of traditional methods that rely on 2D projections or existing shapes. This approach integrates a continuous latent representation with physics-aware optimization strategies, allowing for more detailed and innovative design exploration.

A recent study published in Nature — Machine Learning explores the biologically-informed integration of drug representations for breast cancer treatment using deep learning techniques. This innovative approach aims to enhance the efficacy of treatment strategies by leveraging advanced computational methods to better understand drug interactions and tumor biology.