Machine-Learning and Artificial Intelligence May Learn More Flexibly with New Method

Imagine machines that can learn more flexibly over time. Scientists have created a new way of doing machine learning that enables semantically related concepts to reinforce each other.

"When you have a lot of possible categories, the conventional way of dealing with it is that, when you want to learn a model for each one of those categories, you only use data associated with that category," said Chiyuan Zhang, one of the researchers, in a news release. "It's treating all other categories equally unfavorably. Because there are actually semantic similarities between those categories, we develop a way of making use of that semantic similarity to sort of borrow data from close categories to train the model."

In this latest study, the researchers actually found that a machine-learning algorithm that used their training strategy did a better job of predicting the tags that human users applied to images on the Flickr website than it did when it used a conventional training strategy.

This strategy, for example, would include an object-recognition algorithm that would learn to weight the co-occurrence of the classifications "dog" and "Chihuahua" more heavily than it would the co-occurrence of "dog" and "cat."

In order to perform this type of complicated evaluation, the researchers used a metric called the Wasserstein distance, which is a way of comparing probability distributions. In experiments, the researchers found that their system outperformed a conventional machine-learning system even when the criterion of success was simply predicting the tags that Flickr users had applied to a given image.

The findings may be a huge step up for machine learning. By associating objects with one another, machines can work better and more efficiently.

Related Stories

Robotic Hands May be More Dexterous with New Embedded Optical Sensors

Robot Teaches Us Why Babies Smile So Much (VIDEO)

For more great science stories and general news, please visit our sister site, Headlines and Global News (HNGN).