Abstract: Deep learning refers to Artificial Neural Networks (ANN) with multiple layers. These networks are inspired by human brains and contain billions of neurons like human brain for communication. There are various types of architectures of Neural Networks and among all two of them is Multilayer perceptron and Bayesian Neural Network. Multilayer Perceptron is a feed forward neural network with more than three layers. Whereas, Bayesian Neural Network is an extended version of this Multi-layer perceptron neural network that use Bayes theorem to describe the uncertainty in weights so that uncertainty in predictions can be estimated which are not estimated by simple multi-layer perceptron. In this paper Multilayer perceptron neural network having 7 layers and Bayesian neural network having 7 layers is implemented and compared on Bank Telemarketing dataset. Finally, Accuracy, ROC-AUC curve, Binary cross entropy, and KL-divergence loss are used to compare both the models.
Keywords: Neural Networks, Bayesian Neural Networks, Multi-Layer Perceptron, Bank telemarketing, probability distribution, classification.
|
DOI:
10.17148/IJARCCE.2022.117103
