Computer scientists are developing such technologies that use artificial intelligence which learns or adapt things continuously similar to that of a human brain. Human brain learns new things simultaneously and never gets tired of it, just even thinking about what if one stops to learn? Life will be boring right.
Neural networks are a set of computing elements communicate through the connections that vary in weight these are all forms which machine learning algorithms take just think of an algorithm that is designed to recognize pictures it will set certain values or weights. Scientists have introduced a new technique through which influences one neuron to activate the other more and this happens when the weight is splitter into the furthermore two values. The first value from the splinted ones trained in a traditional system but the second value responds to its surroundings and adjusts itself, algorithms also know how to make these weights or values more adjustable accordingly.
There was a conference held in Stockholm, Sweden in July the researchers and scientists presented their techniques about how to transform the response in new situations and circumstances. The technique was basically to create a network that can create a picture that is half-erased photo even it has seen the full photograph only a few times, while the traditional neural networks aren’t able to do this as they have to see the full image many times. They also introduce a network that can learn how to understand the alphabets which aren’t like the typed ones after seeing them only once.
Another task performed was to control the moving character to find reward by the neural network. achieved after trying it one million times as could a network with only fixed weights. There are two parts of semi adjustable weights static parts and the dynamic parts, static ones apparently learned the structure of maze, whereas the dynamic ones learned how to understand new reward locations.
A computer scientist in the University of California Berkeley, Nikhil Mishra added “This is powerful as the algorithms can adapt quickly to new tasks and new situations, just like humans” although the scientist wasn’t the part of the research. Another computer scientist at ride-sharing company Uber and the papers lead author named Thomas Miconi wanted to challenge more complicated and complex tasks such as robotic control and speech recognition with the help of his teamwork.