Researchers uncovered how deep networks think.
"A revolutionary method finally unlocks the mystery of how AI works. By uncovering the logic behind its decisions, this research shows how machines categorize the world and ensures AI systems are safer and smarter. Credit: SciTechDaily.com" (ScitechDaily, How Scientists Are Finally Revealing AI’s Hidden Thoughts)
Image 2 introduces the deep (learning) neural network. When we look at the neural network we can see that there are four main layers. But when we think that the deep neural network is an AI-based system there is a possibility that the neural network can distribute or divide and scatter that entirety into the multiple sub entireties.
That means there can be multiple different layers. And if every single point is one single computer. That means the computer can take the mission and share it with other computers. This is the reason why the deep network can be so effective.
The thing is that those computers in the deep networks can operate separately or they can operate in entirety. The system can cut problems into smaller parts and then share them with computers. If the deep network has some more urgent mission. It can save the mission. And then handle that more urgent mission. Or the deep network can turn itself into multiple different networks. And all of those networks can operate independently. Then they can connect their results into a new entirety.
Image 2: Deep network.
The deep network emulates human brains. The system thinks through layers. And that makes it a very powerful tool. The deep neural networks can drive different types of information into their layers horizontally. Then deep can compare those datasets with each other. That means the system puts the datasets together vertically.
Or it can connect them with new data and make new connections in those datasets. Deep neural networks can search the differences between two objects or they can find overlaps from those datasets. And that allows the deep neural network to eliminate those overlaps from datasets. And that makes those systems more effective. Because. They must not handle the same information many times.
The ability to connect data into new effective entireties makes the deep networks tools that can beat any single computer. The deep network can connect datasets partially. That means it must not use its all force in the data handling process. If the four-layer system in a deep network uses two of its layers for searching differences. The are two layers for searching for similarities. Or for some other things.
"Introducing the k* Distribution Method" (ScitechDaily, How Scientists Are Finally Revealing AI’s Hidden Thoughts)
"In this study, the researchers developed a new method, called the k* distribution method, that more clearly visualizes and assesses how well deep neural networks categorize related items together." (ScitechDaily, How Scientists Are Finally Revealing AI’s Hidden Thoughts)
"The model works by assigning each inputted data point a “k* value” which indicates the distance to the nearest unrelated data point. A high k* value means the data point is well-separated (e.g., a cat far from any dogs), while a low k* value suggests potential overlap (e.g., a dog closer to a cat than other cats). When looking at all the data points within a class, such as cats, this approach produces a distribution of k* values that provides a detailed picture of how the data is organized."(ScitechDaily, How Scientists Are Finally Revealing AI’s Hidden Thoughts)
“Our method retains the higher dimensional space, so no information is lost. It’s the first and only model that can give an accurate view of the ‘local neighborhood’ around each data point,” (ScitechDaily, How Scientists Are Finally Revealing AI’s Hidden Thoughts)
https://scitechdaily.com/how-scientists-are-finally-revealing-ais-hidden-thoughts/
https://en.wikipedia.org/wiki/Deep_learning
Kommentit
Lähetä kommentti