Chinese scientists have developed a reconfigurable integrated photonic chip capable of running diverse neural networks, including those for image and speech processing, with high efficiency.
Key Details
- 1The chip integrates microring resonators and Mach-Zehnder interferometers, powered by a soliton microcomb light source.
- 2Supports fully-connected, convolutional, and recurrent neural networks within a single hardware architecture.
- 3Area efficiency reaches up to 2.45 TOPS/mm² at 10 GHz frequency.
- 4Demonstrated on tasks: image classification (MNIST 92.93% accuracy, CIFAR-10 56.57%), sentiment analysis (IMDB 80.81%), and speech recognition.
- 5Device enables dual-path computation per resonator, doubling throughput versus traditional schemes.
Why It Matters
This advance shows that photonic computing chips can support the flexible and scalable deployment of multi-model neural networks for imaging and multimodal AI tasks on efficient, high-throughput hardware. Such technology may lead to breakthroughs in hardware acceleration for radiology AI and similar domains.
Source
EurekAlert
Related News
•EurekAlert
AI Repurposes Routine CT Scans for Osteoporosis Detection
AI algorithms can extract bone density data from routine CT scans to identify osteoporosis, enabling opportunistic screening.
•EurekAlert
AI Outperforms Radiologists in Detecting Hidden Airway Objects on Chest CT
Southampton researchers developed an AI that surpassed radiologists in detecting hard-to-see airway obstructions on chest CT scans.
•EurekAlert
AI Method Automates X-ray Absorption Spectroscopy for Material Analysis
Researchers have developed an AI-based approach to automate and enhance the analysis of X-ray absorption spectroscopy (XAS) data for materials science.