Researchers at the École Polytechnique Fédérale de Lausanne (EPFL) Embedded Systems Laboratory (ESL) in Switzerland have developed an artificial intelligence that personalizes and improves video streaming.
“Platforms like YouTube or Netflix use two systems, both of which are inefficient,” Marina Zapater Sancho, researcher at ESL and one of the authors of the study being carried out under the MANGO project, said in a statement. “They store either one copy of a video in the highest-quality format possible, or dozens of copies in different formats.”
According to the ESL researchers, storing one copy of a video in the highest-quality format possible can result in choppy and slow streaming, while storing dozens of copies in different formats eats up lots of power and takes up huge amounts of server storage.
Arman Iranfar, researcher at ESL and one of the authors of the study, said that they have developed a way of reducing power requirement by nearly 20% while improving the streaming experience by 37% through a method that involves machine learning.
“Computers learn from experience,” Iranfar said. “We exposed them to many different scenarios, such as 1,000 people playing a video, each from a different device. The computers remembered the series of actions that led to positive outcomes and reproduced them.”
According to the ESL researchers, systems that encode videos for streaming can similarly learn to personalize the allocation of resources based on series of actions of users while simultaneously optimizing quality, compression, performance and power consumption.
Current video streaming doesn’t allow personalized service. Regardless of the device use, internet connection and the viewing environment, an online video is watched in a similar manner by thousands of people.