Nvidia shrinks LLM memory 20x without changing model weights

Nvidia's KV Cache Transform Coding (KVTC) compresses LLM key-value cache by 20x without model changes, cutting GPU memory costs and time-to-first-token by up to 8x for multi-turn AI applications.

AI model improves flood forecasting with higher accuracy than current methods

New paired studies from the University of Minnesota Twin Cities show that machine learning can improve the prediction of ...
Tech Xplore

Hybrid AI model combines graphs and transformers for real-time traffic forecasts

Urban congestion is a big problem in our cities. It leads to commuter delays and economic inefficiency. More tragically, though, it leads to a million deaths annually worldwide. Research appearing in ...

Clustering-based AI forecasts river water levels using just a few long records

Reliable and scalable water level prediction is crucial in hydrology for effective water resources management, especially ...
IEEE

Leveraging Knowledge Graph and Large Language Model Synergies for Intelligent Fault Analysis in Urban Rail Transit Signaling Systems

Abstract: To address the challenges of insufficient text data utilization and low manual diagnosis efficiency in urban rail transit signaling system fault analysis, particularly the difficulty in ...
IEEE

KKGCN: A Knowledge-constrained K-hop Graph Convolutional Network for Soft Sensor modeling in Industrial Processes

Abstract: With the increasing complexity of industrial processes, the effective monitoring of key quality variables has become critically important. Soft sensing techniques provide an efficient means ...