What are the dogs at home thinking about? Scientists use machine learning to figure it out

Have you ever thought about what is going on in the mind of a dog that acts coquettishly and begs to be fed to you every day?

Why do dogs that put a lot of effort into raising mature dogs just turn around and jump into other people's arms?

Actually, dogs don’t do this intentionally to make you angry -

An experiment from Emory University showed that dogs are The world may see things very differently than we do.

Humans pay more attention to objects, but dogs don’t care much about who or what object they see, but care more about the action itself.

So it stands to reason that anyone who treats a dog well can get close to him. (Of course, don’t forget that dogs have a keen sense of smell, and many dogs also recognize their owners)

In addition, the visual systems of dogs and humans are also very different. They can only see yellow and blue tones, but there is one Sensitive visual receptors used to detect movement.

This experiment used machine learning to figure out the dog’s brain activity, and the related paper has been published in The Journal of Visualized Experiments.

Researchers pointed out that this method has an obvious advantage: it is non-invasive.

This method has only been used on primates before, so this experiment on dogs is a major breakthrough.

Let’s take a look at the specific experience process.

Comparing the brain activity of dogs and humans

Researchers used machine learning and fMRI (functional magnetic resonance imaging) to explore the brain activity of dogs when they watch different types of videos.

As for why you should choose dogs instead of other animals?

Because dogs are relatively easy to teach animals, after certain training they can obediently cooperate with MRI (magnetic resonance) scans without the need to inject sedatives or use other restraint methods.

However, although the dogs are relatively obedient, in this study, in addition to participating in MRI scans, they also had to watch videos for a long time. So in the end, only two dogs were selected, one was a 4-year-old male Boxer mix and the other was an 11-year-old female Boston Terrier mix.

△The dog is watching the video

Each dog watched three different sets of videos, each set of videos lasted 30 minutes, a total of 256 videos fragment. To control for variables, these videos had no sound.

Some of the videos focus on showing different objects (e.g. people, dogs, cars), others focus on showing different actions (e.g. playing, eating, smelling).

For comparison, two human volunteers also watched the video clips using the same procedure.

While the volunteers and dogs watched the video, the researchers used a 3T MRI scanner to record images of their brain activity.

They then used neural networks to train and test three classifiers (classifiers) to distinguish between "objects" and "actions".

Hey, why not 2 but 3?

Because among the behavior classifiers, one was trained on 3 different actions, and the other learned 5 actions.

The results show that the human brain responds well to objects and actions, while the dog's brain is only sensitive to actions. They don't seem to be cold to different people and objects.

△The left is the human brain MRI image, and the right is the dog brain MRI image

In order to evaluate the performance of the model and make the data more convincing, the researchers also used The Ivis machine learning algorithm is used to quantify the collected data.

Looking at data from human volunteers, these models all achieved over 99% accuracy in mapping their brain activity data to different classifiers.

When decoding dog brain data, the model is basically ineffective for object-based classifiers; however, for behavior-based classifiers, the accuracy can reach 60% to 88%.

It can be seen that the way dogs think is very different from ours!

About the researcher

The researcher is from the Department of Psychology at Emory University.

The first author of the paper, Erin M. Phillips, is currently a doctoral candidate in the Department of Ecology and Evolutionary Biology at Princeton University.

She came to Emory University as a visiting scholar and participated in this research.

The co-author of the paper, Gregory S. Berns, is currently a professor at Emory University. His main research directions are neuroimaging of human decision-making, dog fMRI, and comparative neurobiology.

Professor Berns graduated from the Department of Physics of Princeton University with a bachelor's degree and received two doctorate degrees: a doctorate in biomedical engineering and a doctorate in medicine.

Berns believes that it is understandable that dogs pay more attention to movements, because animals naturally need to pay close attention to changes in the environment in order to hunt or avoid being eaten.

Although only two dogs participated in this study, the researchers will conduct experiments on more dogs and other animals in the future to explore how animals perceive the world.

Paper address: https://www.jove.com/t/64442/through-dog-s-eyes-fmri-decoding-naturalistic-videos-from-dog

Reference link ：https://www.eurekalert.org/news-releases/964886

The above is the detailed content of What are the dogs at home thinking about? Scientists use machine learning to figure it out. For more information, please follow other related articles on the PHP Chinese website!