I forwarded this to one of the statisticians, who replied:
***
By default the procedure uses a single hidden layer with automatic choice of the number of units in the hidden layer, with between 1 and 50 units. The algorithm for automatic selection is not simple. It has a random component in that it splits the data into training and test samples in order to do the selection.
It's very simple to specify that you want one or two hidden layers, and the number of units in each. On the Architecture tab, select Custom architecture, one or two layers, and under Number of Units, again Custom, and specify the number desired in each hidden layer. This controls syntax for the ARCHITECTURE subcommand.
Note that the number of units in each layer does not include the bias.
***
------------------------------
Rick Marcantonio
Quality Assurance
IBM
------------------------------
Original Message:
Sent: Wed December 07, 2022 04:03 AM
From: jamal numan
Subject: IBM SPSS 26: How to specify the number of perceptron (neurons\nodes\junctions) in each hidden layer in the multilayer perceptron (MLP)?
IBM SPSS 26: How to specify the number of perceptron (neurons\nodes\junctions) in each hidden layer in the multilayer perceptron (MLP)?
I couldn't figure out how to specify the number of perceptron (neurons\nodes\junctions) in each hidden layer in the multilayer perceptron (MLP). Is the number of neurons randomly determined?
For example, why the number of neurons in the MLP below is 2?
------------------------------
jamal numan
------------------------------
#SPSSStatistics