One of the key methods utilised in the area of artificial intelligence is pattern recognition. The most sophisticated senses in the human body, vision and perception, served as the inspiration for the computer paradigm known as neural networks. All real-time applications revolve on pattern recognition. Human decision-making is influenced by our ability to recognise patterns. Neural network classifiers are created for pattern recognition applications in this research effort. For the purpose of evaluating the benefits and drawbacks of the suggested Neural Network Classifiers, the job of handwritten character recognition has been selected as the application domain. With the aim of achieving rapid training, the standard Backpropagation method for Feedforward Neural Networks is changed. The proposed adaptive backpropagation method dynamically modifies the learning constants based on adaption cycles and error. When compared to traditional methods, the error is observed to reduce more quickly. The categorization of the Iris data set and the identification of handwritten characters serve as proof of the algorithm's effectiveness. Learning and generalisation are greatly influenced by the network size. The ideal network size is chosen using adaptive genetic algorithms. Depending on how sensitive the edges are to mistake, duplicated connections that contribute little to nothing are cut.

Neural Network, Machine Intelligence, Pattern Recognition, Conventional Algorithms and Backpropagation