The Defense Advanced Research Projects Agency (DARPA) is seeking industry assistance to develop efficient analog neural networks for sensor processing in constrained spaces. The Scalable Analog Neural-networks (ScAN) project aims to create neural networks that can directly interface with conventional sensor outputs, resulting in a significant reduction in power consumption. This initiative showcases DARPA’s commitment to enhancing artificial intelligence systems for military applications.
Analog neural networks mimic the human brain’s data processing capabilities, enabling computers to learn from their mistakes and improve continuously. The ScAN program intends to demonstrate the inferencing capabilities of analog neural networks while eliminating the need for analog-to-digital converters at the sensor level. By leveraging analog technology, DARPA aims to increase the power efficiency of neural networks while ensuring accuracy, robustness, and scalability.
Traditional digital neural networks often face power consumption limitations, making them unsuitable for applications with restricted sensor processing requirements. The ScAN program seeks to overcome these challenges by utilizing analog technology to enhance power efficiency and performance reliability. This initiative emphasizes the development of new analog neural network processing architectures and algorithms to address current limitations in digital neural network systems.
Companies interested in participating in the ScAN program must submit abstracts by July 8, 2024, and full proposals by August 8, 2024, through the DARPA Broad Agency Announcement Tool (BAAT) online platform. DARPA encourages the use of existing computing infrastructure or cloud-based services for all computing needs to streamline the development process. By fostering collaboration between industry and government agencies, the ScAN program aims to revolutionize sensor processing capabilities for military applications.
