Small data train ANN

I have small data input matrix [I N]=[4 26] targets [O N]=[1 26] .Is it correct to train all data with dividetrain (val=test=0) and then test to unseen?Per example train in 18 of 26 and test in 8 (unseen) ?How can i approach this classification problem?

Answer 1

[Greg Heath]

Use k-fold Cross validation for a range of k values and multiple random initial weights for each value of k..

Thank you for formally accepting my answer

Greg