Is it possible truncate a neural network weights?

When I am training a neural network I get the weights e.g. 0.960928653806898,0.273172876012661, ..., -0.413223736485805, but is possible modify the weights to obtain only the integer part and 2 digits to the fracctional part?

Answer 1

[Greg Heath]

Set the display format at the beginning of the program. It will not affect the accuracy of internal function calculations.

 help format
 doc format

e.g.,

 a = 0.960928653806898; b=0.273172876012661; c=-0.413223736485805;
 format short
 a,b,c

 %a =  0.96093
 %b =  0.27317
 %c = -0.41322

Hope this helps.

Thank you for formally accepting my answer

Greg

[Image Analyst]

Ana's answer moved here (It seems like Ana and Shana are the same person with two accounts):

Hi Greg, thank you for answer my question. At the beginning of the program I write "format compact" and with this command I visualize weights like: 0.9609,0.2731, ..., -0.4132. However, I know matlab weights are obtained randomly. But is possible truncate the randomly number using a function like round2 to obtain weights like 0.96,0.27, ..., -0.41?