When you request a ride from an app, the company’s computers work to match you with a driver quickly. They consider that you need a fast ride, you’re not the only one needing a ride, and drivers want to pick up nearby passengers. This process, called bipartite matching, is also used for pairing organ donors with recipients, medical students with residencies, and advertisers with ad slots.
Cold Spring Harbor Laboratory Associate Professor Saket Navlakha found a better way by looking at biology. Navlakha saw that muscle fibers are first connected to many neurons in the nervous system but need to end up with just one.
Neurons compete to stay connected to the same muscle fiber using neurotransmitters like bids. Neurons that lose the bid move to other fibers, ensuring every neuron and fiber eventually pair up correctly.
Navlakha created a simple algorithm based on nervous system matching. It uses two equations: one for neuron competition and one for reallocating resources. The algorithm performs well, creating nearly perfect pairings and reducing unmatched pairs. It could improve wait times for rideshares and hospital staffing. Plus, it protects privacy by avoiding central servers.
Navlakha hopes others will use this algorithm for various applications, showing how studying neural circuits can inspire new AI solutions.
The study shows that we can create a simple and effective algorithm by mimicking the nervous system’s way of matching neurons to muscle fibers. This approach improves pairing efficiency, reduces unmatched pairs, and protects privacy.
It offers a promising method for solving various real-world problems. It highlights how studying the nervous system can lead to innovative technological solutions.
Journal reference :
- Dasgupta, S., et al., “A neural algorithm for computing bipartite matchings. PNAS. DOI: 10.1073/pnas.2321032121.
