Nadkarni, Devika2025-06-162025-06-162014https://hdl.handle.net/2144/50673Artificial neural networks are computing systems inspired by the processing connections of neurons in the mammalian cerebral cortex. Regular computers process serially, in single step by step sequences, with various approximations such as normality and linearity. Neural networks differ from the traditional computing architecture in that they process information in parallel through multiple, interconnected simple units. In the human brain, these elementary units are neurons, and number in about 87 billion. The connections between these billions of neurons form the basis of our unique processing abilities. In networks, each unit may receive input and send output to multiple other neurons rather than just one. This creates a branched network capable of processing information in parallel - that is, multiple sequences of information processing can be created rather than a single sequence. In comparison to traditional computing systems, networks modeled after the brain tend to be slower and less precise. However, what they lack in speed is compensated by their ability to process complex information. Neural networks have value in being able to discern nonlinear, dynamic relationships, unlike serial networks.en-USArticle