Okinawa Institute of Science and Technology develops fly-inspired scent algorithm

Researchers at the Okinawa Institute of Science and Technology (OIST) have developed a new energy-efficient algorithm inspired by the olfactory system of the fruit fly. Published in Neuromorphic Computing & Engineering, the system aims to address the challenge of creating functional, low-memory computational solutions that can operate in real-world scenarios where energy is limited and training data arrives progressively.

The algorithm mimics the fruit fly's brain, which contains approximately 140,000 neurons and utilizes sparse coding to reduce signal overlap without extensive training. This method allows for the efficient representation of scents with minimal energy costs. In testing, the algorithm successfully identified odors with high accuracy, demonstrating particular strength when working with limited sets of training data.