Researchers at Intel Labs and Cornell University have utilized an unusual “neuromorphic chip” to quickly learn the signature smell of ten different hazardous chemicals and spot their presence quicker than ever before. The Loihi chip, as it is called, mimics how our brains classify and identify unique smells that our noses detect, retaining a memory of a particular chemical after only a single exposure to its smell. Existing technologies that perform similar feats require hundreds or thousands of exposures to a chemical before finally learning to identify it consistently.
“We are developing neural algorithms on Loihi that mimic what happens in your brain when you smell something. This work is a prime example of contemporary research at the crossroads of neuroscience and artificial intelligence and demonstrates Loihi’s potential to provide important sensing capabilities that could benefit various industries,” said Nabil Imam, a scientist at Intel’s Neuromorphic Computing Lab.
The team took data gathered from 72 chemical sensors and fed it into the Loihi chip that was adapted to recreate the brain circuits that perform olfactory-related functions. The chip learned on-the-fly and was able to sense and identify test samples right after learning from initial study samples.
Proving that the new chip can be used in challenging applications, its recognition of one of the ten chemicals was successfully achieved even while it was subjected to significant noise and occlusion. As it learned to recognize one chemical, the chip retained a memory of the chemicals it previously encountered, allowing it to expand its capabilities. Compared with a deep learning system previously created, it required 3,000 times fewer training samples to perform at a similar level.
From the study abstract in Nature Machine Learning:
We present a neural algorithm for the rapid online learning and identification of odourant samples under noise, based on the architecture of the mammalian olfactory bulb and implemented on the Intel Loihi neuromorphic system. As with biological olfaction, the spike timing-based algorithm utilizes distributed, event-driven computations and rapid (one shot) online learning. Spike timing-dependent plasticity rules operate iteratively over sequential gamma-frequency packets to construct odour representations from the activity of chemosensor arrays mounted in a wind tunnel. Learned odourants then are reliably identified despite strong destructive interference. Noise resistance is further enhanced by neuromodulation and contextual priming. Lifelong learning capabilities are enabled by adult neurogenesis. The algorithm is applicable to any signal identification problem in which high-dimensional signals are embedded in unknown backgrounds.
Study in Nature Machine Intelligence: Rapid online learning and robust recall in a neuromorphic olfactory circuit