How can I design multi layer perceptron based neural network(2 hidden layers) for my data.

Data consists of 72 features as columns,2148 rows, and last column as class labels as 0 and 1.

[Greg Heath]

One hidden layer is sufficient.

Greg

Answer 1

[Raunak Gupta]

Hi,

Since you are trying to implement a Multi-Layer Perceptron, It is recommended to use fullyConnectedlayer as this connect all the nodes in one layer to the previous and next layer. The input size can be specified as 72 which is the number of features. The rows in the data matrix can be split into test data and train data. For full workflow, there is an example here by ignoring convolution2dLayer, maxPooling2dLayer and mentioning imageInputLayer with appropriate size as the number of features.

You may use Neural Netwrok Pattern Recognition App as an interactive alternative for solving classification and regression-based problem in Neural Networks.

For more information you may take a look at Neural Network.

[S NISHANTH]

ok thank u