If you pick up a grapefruit and take a whiff, the fruit’s molecules stimulate olfactory cells in your nose. Those cells immediately send signals to your brain, where electrical pulses within an interconnected group of neurons generate a smell’s sensation.

Whether you’re smelling a grapefruit, a rose, or a noxious gas, networks of neurons in your brain create sensations specific to the object. Similarly, your senses of sight and sound, your recall of memory, your emotions, and your decision-making each have individual neural networks that compute in particular ways.

But we humans aren’t the only ones who can learn to identify by smell. Researchers from Intel Labs and Cornell University have developed mathematical algorithms that allow a neuromorphic research chip to rapidly learn neural representations of 10 different odors.

The self-learning chip, Loihi, mimics how the human brain functions by learning to operate based on various modes of feedback from the environment. The researchers trained Loihi to recognize hazardous chemicals, even in the presence of significant noise.

Since Loihi gets smarter over time and doesn’t need to be trained in the traditional way, it demonstrated superior recognition accuracy compared with conventional methods. For example, a deep learning solution to recognize odors required 3000x more training samples per class.

“My friends at Cornell study the biological olfactory system in animals and measure the electrical activity in their brains as they smell odors,” explains Nabil Imam, a senior research scientist in Intel Labs’ neuromorphic computing group. “On the basis of these circuit diagrams and electrical pulses, we derived a set of algorithms and configured them on neuromorphic silicon, specifically our Loihi test chip.”

Detecting distinct odors

Imam and his team took a dataset consisting of the activity of 72 chemical sensors in response to 10 gaseous substances (odors) circulating within a wind tunnel. The sensors’ responses to the individual scents were transmitted to Loihi where silicon circuits mimicked the circuitry of the brain underlying the sense of smell.

The chip rapidly learned neural representations of each of the 10 smells, including acetone, ammonia, and methane, and identified them even in the presence of strong background interferents. Your smoke and carbon monoxide detectors at home use sensors to detect odors but they cannot distinguish between them; they beep when they detect harmful molecules in the air but are unable to categorize them in intelligent ways.

The chemical-sensing community for years has looked for smart, reliable and fast-responding chemosensory processing systems, otherwise called “electronic nose systems.”

Imam sees the potential of robots equipped with neuromorphic chips for environmental monitoring and hazardous materials detection, or for quality control chores in factories. They could be used for medical diagnoses where some diseases emit particular odors. Another example has neuromorphic-equipped robots better identifying hazardous substances in airport security lines.

“My next step,” Imam says, “is to generalize this approach to a wider range of problems — from sensory scene analysis (understanding the relationships between objects you observe) to abstract problems like planning and decision-making.”

“Understanding how the brain’s neural circuits solve these complex computational problems will provide important clues for designing efficient and robust machine intelligence,” he says.

But there are challenges in olfactory sensing. When you walk into a grocery store, you might smell a strawberry, but its smell might be similar to that of a blueberry or a banana, which induce very similar neural activity patterns in the brain.

Sometimes it’s even hard for humans to distinguish one fruit from a blend of scents. Systems might get tripped up when they smell a strawberry from Italy and one from California, which might have different aromas, yet need to be grouped into a common category.

“These are challenges in olfactory signal recognition that we’re working on and that we hope to solve in the next couple of years before this becomes a product that can solve real-world problems beyond the experimental ones we have demonstrated in the lab,” Imam says.

His work, he contends, is a “prime example of contemporary research taking place at the crossroads of neuroscience and artificial intelligence.”

Read more:

• Building a better brain
• The next big buzz in AI
• Nose in the sky