Latest News

A team from the University of Copenhagen’s Department of Computer Science — Dustin Wright, Christian Igel, and Raghavendra Selvan — has introduced a novel approach to making deep learning models more efficient without compromising accuracy. Their method, called Bayesian Model Reduction for Structured Pruning (BMRS), brings a principled Bayesian perspective to one of the field’s most pressing challenges: reducing the massive computational cost of modern neural networks.

Researchers Mengxi Liu, Sizhen Bian, Zimin Zhao, Bo Zhou, and Paul Lukowicz from the German Research Centre for Artificial Intelligence (DFKI) have introduced an innovative approach to real-time eye blink detection using capacitive sensing technology. Their paper, “Energy-efficient, low-latency, and non-contact eye blink detection with capacitive sensing”, presents a lightweight, wearable prototype embedded into a pair of standard glasses—eliminating the need for cameras or skin contact.

This work reflects DFKI’s continued commitment to developing sustainable, privacy-respecting, and user-friendly sensing solutions for human–computer interaction, wearable computing, and assistive technologies.

The SustainML Framework has extended its metadata management capabilities by integrating a new module for automated dataset profiling and semantic description generation using Large Language Models (LLMs). This development enables automatic generation of dataset metadata — including descriptions, semantic classifications, and contextual insights — directly from raw tabular or structured data.

Researchers from the German Research Center for Artificial Intelligence (DFKI) have released a timely and critical study on how irregular sampling in time-series data affects AI performance in Human Activity Recognition (HAR). Presented at the International Conference on Pervasive Computing and Communications (PerCom), the paper "Assessing the Impact of Sampling Irregularity in Time Series Data: Human Activity Recognition as a Case Study" exposes limitations in both discrete-time and continuous-time neural networks when faced with real-world sensor noise, packet loss, and jitter.