The portable device they have created at the University of California in Los Angeles allows to identify allergens in the air based on deep learning.
At the University of California at Los Angeles (UCLA) they have created a device of an everyday utility. Now that every time allergies appear more frequently in the most unexpected moments, a team of scientists has come up with a portable solution that can come in handy for many who suffer from these types of conditions.
Every minute, an adult breathes between 100 and 1,000 bioaerosols (aerosols formed by particles of biological origin). These include pollens, spores, toxins, and microbes. If this person lives in a highly polluted environment, the figure amounts to one million bioaerosols. This type of particles are what cause allergies, in addition to causing asthma and other diseases.
The UCLA team of scientists has managed to create a portable device that combines image capture with deep learning. Its purpose: to measure bioaerosols, which originate from living organisms such as plants and fungi, and to identify allergens in the air.
Much of the utility of the device is that it can be easily moved, so that people with allergies can measure air quality in different areas. It does so because it is trained to recognize five types of allergens in the air. His hit rate is 94.% , according to the deep learning system on which it is based.
Everything is compare
Since the concept of artificial intelligence began to be used in a practical way, there is a basic behavior that underlies everything. It’s about comparing. Record, measure, check. They are actions that in the end are equivalent to taking a sample and comparing it with some parameters that we already have classified and ordered.
This has been the case ever since and continues to be the case, despite the increased capacity involved in machine learning and, now, deep learning neural networks. The method used by the device created in UCLA is no exception. Your deep learning is based on take particles from the air and transform them into digitally actionable information.
From this data, already translated into bits, an algorithm cuts the information, actually an image, and represents the biological particles. Another second algorithm, based on a neural network, classifies these particles based on a cataloguing of different types of allergens, a database that the system counts on to compare.
In the end it is a comparison between new information and a perfectly classified and ordered database. And much of the success it really depends on the success when it comes to labeling the elements from our database.
Images: Coley Christine, Brittany Colette
