Echo state networks (ESNs) are a promising machine learning framework for predicting changing patterns over time, but their accuracy depends strongly on how their settings are chosen. Researchers at Tokyo University of Science found that the time scale of a target system, or how quickly it changes, can help guide these settings. The findings could make ESNs easier to tune and improve applications such as weather forecasting, robotics, and signal processing.

Detecting microcystin-leucine-arginine (MC-LR) in freshwater is increasingly important as harmful algal blooms intensify, but biosensor readings are affected by changing water conditions. Researchers at Hanbat National University and the University of Central Florida developed a machine learning-based calibration framework for biosensors using water quality data. The model accurately predicted MC-LR levels while reducing the need for repeated recalibration, enabling more reliable on-site monitoring.

CDE research expertise in environmental resilience and turning complex waste materials into reusable resources will play a central role in a new national initiative to advance waste innovation in Singapore and build a more resource-efficient future.

Launched on 17 June by Senior Minister of State, Ministry of Sustainability and the Environment, Dr Janil Puthucheary, at Catalyst 2026, the Towards Resource Efficiency And Sustainability for URban EnvironmentS (TREASURES) centre will be Singapore’s first national research centre dedicated to advancing science and innovation in residue and toxic industrial waste management.