Artificial neural networks are modelled on the human brain and nervous system. They are used to estimate or approximate functions that can depend on a large number of inputs that are generally unknown. One of the hottest areas of artificial intelligence development, neural networks are used extensively in image recognition, data science, forecasting, and many more.
Ed Tyantov, one of the developers of Russian neural network-powered app Artisto, has used neural networks to determine who most looks like a US president. The answer he got was Donald Trump.
The developer used the model created by Mu Li, a student at the University of Carnegie Mellon (Pittsburgh, PA), for the Large Scale Visual Recognition Challenge 2016. After analyzing a photo, the model provides five most relevant characteristics for the person or object pictured.
Li’s neural network was trained using a set of 14 million categorized images, each image was assigned to one of 21,000 categories (for example, a profession, position, etc.). The network achieved prediction accuracy of 70% for the five most relevant characteristics.
During the learning process set up by Tyantov, the neural network in question has “seen” more than a thousand photos of various heads of state. As a result, it has learned how presidents look. It has learned to recognize facial features, clothes, accessories, background pictures, etc. typical for presidents.
So, what did the neural network say about the man in this photo?
The system provided the following characteristics:
- Ex-president
- President
- Secretary general
- President of the United States, president, executive director,
- Minister
And here is how the neural network characterized Hillary Clinton:
- Secretary of State
- Donna
- First lady
- Auditor
- Girl
Photos of Vladimir Putin, Barack Obama and Angela Merkel were also used.
The neural network identifies the current US president as “the head of the company, chief operating officer,” “minister,” “president of the United States executive director,” “executive vice president,” “baron, big businessman, king, magnate, business leader, top manager.”
The network calls Vladimir Putin “centrist, moderator,” “president,” “president of the United States, head of the company,” “chief executive officer, chief operating officer,” “former president.”
Angela Merkel is characterized as “secretary of the state,” “minister,” “executive vice president,” “skeptic,” “chief executive officer, head of the company, chief operating officer.”
Tyantov’s open source model for president photos is available here.
Source: [email protected]