Introduction In this essay we are going to discuss
In this essay we are going to discuss, how a team of researcher has developed a machine learning technique that can exactly reconstruct images transmitted over Optical Fibers for distance up to a kilometer, said monthly journal Optica. This approach could improve medical Diagnostic and telecommunications.
Machine-Learning Algorithm to Boost Telecom
The machine learning technique researchers has reported that teaching a type of machine learning algorithm known as a Deep neural network(DNN) is used to recognize images of number from the pattern of spots they created when transmitted to a fiber. This research can improve the telecommunication networks and endoscope imaging in medical diagnosis and optical power delivered by fiber. The deep neural network is designed to get back the input images from a scrambled output of the fiber “said Demetri Psaltis, Swiss Federal Institute of Technology, Lausanne along with Christophe Moser concluded that optical fiber can last up to a km long & named it as “An important milestone”.
The Optical Fiber carry information in a form of light to all the (MMF) channels which are human eye finds difficult to understand to solve this problem Psaltis & his team used the deep neural network(DNN) a type of machine learning algorithm that work same as a human brain does for example the input processed through multiple layers of artificial neurons conduct small Calculation & passed it on to next layer. The machine identifies the input by recognizing the pattern of output associated with it. The neural networks functions like human brain. The former student Eirini kakkava gave her point of view to explain about the deep neural network functioning by an example” when a person look at an object, neurons in the brain start functioning and recognize the object our brain can do this because it gets trained throughout our life with images or signals of the same category of objects, which changes the strength of the connection between the neurons”. To train an artificial neural network researcher follow essentially the same process, the research is as follow.
To train their system, the researchers, used a database of 20,000 samples of handwritten number zero through nine. Using 16000 as a training data, they kept aside 2,000 for validation and the remaining 2,000 to test the validation system while the spot collected for each digit looked the same to the human eye, But the neural network could differentiate the intensity patterns associated with each digit. Another Researcher Navid Borhani said that machine learning is same as the other method used for reconstructing images through Optical Fiber. The DNN was able to deal with the problem or issue caused by environmental disturbances to the fiber such as temperature fluctuations and the problem caused by slight movement because of Breaking or circulation can distract the image transmitted through & multimode fiber said professor Demetri Psaltis, deep neural networks are a promising solution, for correcting noise caused by these distortions.
The Researcher are trying to develop a new technique with biological samples and overcome the problem faced by noise distortions and to explore the possibilities and limits of their techniques.