Artificial neural networks: an overview and applications
KeywordsPerceptron ; Backpropagation ; Breast cancer diagnosis ; Optical character recognition ; Artificial neural networks ; Neural networks ; Machine learning
This dissertation describes the concept of Artificial Neural Networks for Cancer Prognosis and Handwritten Digits Recognition. Artificial Neural Networks inspired by biological neural networks, are efficiently used to model complex relationships between input signals and outputs. When analyzing the detection of cancer cells the model chosen is a Multi-Layer Perceptron, consisted of a variable number of hidden layers, trained using a back-propagation algorithm. Taking into consideration the characteristics of cell nucleus, provided by Breast Cancer Wisconsin dataset, the model can accurately classify the incident between benign and malignant. Using various different settings of parameters and activation functions our model achieved 97.35% accuracy, showing that it can provide an equivalent or even better alternative to human diagnosis. For the handwritten digits recognition task, the MLP trained using back-propagation algorithm achieved 93.9% accuracy. The noticeable is that the model is having difficulty to distinguish digits that often confused by naked eye.