A new technology has been developed by the engineers with memristors through which the efficiency of teaching machines could be improved in order to think like a human and that is neural network. The words could be predicted before they are said during conversation by this network which is called a reservoir computing system as well as it helps in predicting future outcomes which has been based on the present.
This report has been published in many news papers by the research team that created the reservoir system. Reservoir computing systems have been created in the past with larger optical components which could improve on a typical neural network’s capacity and decrease the required training time.
This system has been created by using memristors by the group of researchers which require less space and it could be integrated more easily into existing silicon based electronics. Memristors could perform both logic and store data as this is a special type of resistive device.
These have been in contrasts with typical computer systems rather logic separate from memory modules has been performed by processors. A special memristor has been used by the team of engineers that memorizes events only in the near history. Neural networks have been composed of neurons or nodes and synapses which have been inspired by brains as there are connections between nodes.
A neural network has taken in a large set of questions and answers to those questions for training a neural network for a task. The connections between nodes have been weighted more heavily or lightly for minimizing the amount of error in getting the correct answer in this process of supervised learning. A neural network could be tested without knowing the answer if these networks once trained.
For example: a new photo can be processed and a human face can be identified by a system as the features of human faces have been learned by it from other photos in its training set. According to the researchers, image recognition is also a simple problem which is related to this and this does not require any kind of information different from a static image.
More difficult tasks could depend highly on context like speech recognition and it demands for neural networks in order to have knowledge of what has been occurred or what has been said. The meaning of a word and pronunciation also would be differ which is depending on the previous syllables at the time of transcribing speech to text or translating languages.