"Neural networks generalized inputs allow you to associate a specific result, "says Ayanna. "When a person sees and hears all fours barking (income), her experience allows us to conclude that this is a dog (the result)." This feature of neural networks enable a new robot select behaviors or paths to follow according to general elements of their surroundings, much like what humans do.
To do this, neural networks contain several layers of "nodes," which are analogous to neurons. Each node in a layer is connected to nodes in other layers. The signals travel in this network of connections, and each node acts as a gate, letting only signs of some magnitude. The network "learns" by adjusting the threshold for each individual node.
Above: In this simplified example of a neural network, input signals are connected to the yellow layer on the left, pass through two layers of process, and then emerge on the right as signs output. This architecture can run surprisingly sophisticated logic problems, especially when feedback loops are added.
This graph of neural networks, which can be drawn on a napkin, it may seem very simple, but in practice, these artificial brains can run incredibly complex logic problems. Ayanna neural networks called "black box technology" - in other words, what happens between the input layer and the output is so difficult to decipher which scientists treat it as a "black box" in a way or other converts inputs into outputs.
Combining these two technologies, Ayanna and her colleagues at JPL expect to create a "brain" robot which alone can learn to walk on unfamiliar terrain of other planets.
To do this, neural networks contain several layers of "nodes," which are analogous to neurons. Each node in a layer is connected to nodes in other layers. The signals travel in this network of connections, and each node acts as a gate, letting only signs of some magnitude. The network "learns" by adjusting the threshold for each individual node.
Above: In this simplified example of a neural network, input signals are connected to the yellow layer on the left, pass through two layers of process, and then emerge on the right as signs output. This architecture can run surprisingly sophisticated logic problems, especially when feedback loops are added.
This graph of neural networks, which can be drawn on a napkin, it may seem very simple, but in practice, these artificial brains can run incredibly complex logic problems. Ayanna neural networks called "black box technology" - in other words, what happens between the input layer and the output is so difficult to decipher which scientists treat it as a "black box" in a way or other converts inputs into outputs.
Combining these two technologies, Ayanna and her colleagues at JPL expect to create a "brain" robot which alone can learn to walk on unfamiliar terrain of other planets.
0 comments:
Post a Comment