summary: Scientists have discovered how animals distinguish between different odors, even when they seem very similar.
Some neurons consistently discriminate between different odors, while others respond unpredictably, helping to distinguish between subtle scents over time. This discovery was inspired by previous research in Drosophila and has the potential to enhance machine learning models.
By introducing variability, AI has the potential to mirror the discriminatory powers found in nature.
Important facts:
- The study discovered two types of neurons: “trustworthy cells” that identify unique odors, and “untrustworthy cells” that help distinguish between similar odors through experience.
- The fluctuations in neural responses were found to come from deeper circuits in the brain, suggesting that they serve an important purpose.
- This neural fluctuation could benefit AI continuous learning systems and increase the system’s insight.
sauce: CSHL
When you order a wine at a fancy restaurant, the sommelier might describe the aroma as citrus, tropical fruit, or floral. Still, when you smell it, you might notice a wine-like aroma. How can wine lovers find such similar scents?
Cold Spring Harbor Laboratory (CSHL) associate professor Saket Navlakha and Salk Institute researcher Shyam Srinivasan may have the answer. They discovered specific neurons that allow fruit flies and mice to distinguish between different odors.
The researchers also observed that, with experience, another group of neurons helps the animals distinguish between very similar odors.
The study was inspired by the work of former CSHL assistant professor Glenn Turner. A few years ago, Turner noticed something strange. When exposed to the same scent, some Drosophila neurons fired consistently, while others varied from trial to trial.
At the time, many researchers dismissed these differences as products of background noise. But Navlaha and Srinivasan wondered if this variation had a purpose.
“There were two things we were interested in,” Navlaha says. “Where does this variation come from? And is there something good about it?”
To address these questions, the team created a fruit fly odor model. The model showed that this variation stems from deeper circuits in the brain than previously thought. This suggests that this variation is indeed meaningful.
The researchers then observed that some neurons responded differently to two very dissimilar odors, but had the same response to similar odors. The researchers called these neurons high-confidence cells. This small group of cells helps flies quickly distinguish between different odors.
Another, much larger group of neurons responds unpredictably when exposed to similar odors. These neurons, which researchers call “unreliable cells,” could help people learn to identify specific aromas in a glass of wine, for example.
“The model we developed shows that these unreliable cells can be useful,” Srinivasan says. “But it takes a lot of learning to take advantage of them.”
Of course, this study isn’t just aimed at wine lovers. Srinivasan said the results explain how we learn to distinguish between similarities detected by our other senses and how we make decisions based on those sensory inputs. states that it may be helpful.
This discovery could also lead to better machine learning models. Unlike neurons in fruit flies or mice, computers typically respond in the same way to the same inputs.
“You may not want your machine learning model to represent the same input in the same way every time,” Navlakha explains. “In a more continuous learning system, variability can be useful.”
That means this research could one day help make AI more discriminatory and reliable.
About this olfactory and neuroscience research news
author: samuel diamond
sauce: CSHL
contact: Samuel Diamond – CSHL
image: Image credited to Neuroscience News
Original research: The survey results are displayed below PLOS Biology
