Question

2. (20) Design an artificial neural network with two hidden layers. First hidden layer has s neurons, second hidden layer has 3 neurons. Input parameters are 3, output parameter i s (20) What is the fundemental philosophy in backpropagation training algorithm, Explain detail. 4 (30) Define the following terms and their effects on the performance of ANN. a) Learning factor b) Momentum factor. c) Number of hidden neuron d) Training data e) Initial Weights Target Output

Answer 1

The learning rate affects a larger or smaller portion of the respective adjustment to the preceding weight. The neural network is easy to learn fastly, if factor is high. Even it is not good at all. For small value, learning rate will be slow. It will execute as a low-pass filter to set abruption changes in the progress. It generally allows a change to the weight to no. of adjustment cycles The degree of persistent is controlled by movement factor. For a non-zero value, greater persistence for current modification Increasing no. of hidden neurons increase the performance of model using MNIST dataset and increase the complexity of network, and is found to be consultable manifested by both visual and mathematical findings Training set characteristics can have a significant effect. Including a prior probability of class relationship into discriminative analysis. Classification accuracy increased and no. of training case get increased \r\n It is one more hyper parameter