Dave, I didn't follow exactly what you are trying to do, but if you can avoid deepcopy() that is probably a good idea.
If you just want to clone a network, you might take a look at the arrayify functions. It assumes that you have the correct structure (layer sizes and connections) and will take all of the learnable network data (weights and biases), turn it into an array which you can then load back into another network. Other than that, you'll probably want to make your own method that would clone the network structure and sizes of layers. -Doug ----- Dave Zoltok <[EMAIL PROTECTED]> wrote: > Hi everyone! > > I'm working on a dissertation involving evolved neural controllers for > robots. I have my experiment running on my own computer, but on my > school's computers, I get the traceback shown below. Each robot is > meant to react to one of several sets of input based on the output > value of a "threshold" neuron. I'm accomplishing this by making a > deepcopy() of the network for each possible input, running them, and > comparing the network output to determine which sets of input I should > pass into the actual propagation. When I try this on the school's > servers, I get this: > > Traceback (most recent call last): > ... > File "/usr/lib64/python2.4/copy.py", line 336, in _reconstruct > y = callable(*args) > File "/usr/lib64/python2.4/site-packages/Numeric/Numeric.py", line > 539, in array_constructor > x.shape = shape > ValueError: total size of new array must be unchanged > > Where should I be looking for the bug here? In Numeric, copy.py, or > the Python install itself? Is there a recommended way of copying the > neural networks in Conx? > > Thanks in advance > Dave > > > _______________________________________________ > Pyro-users mailing list > [email protected] > http://emergent.brynmawr.edu/mailman/listinfo/pyro-users _______________________________________________ Pyro-users mailing list [email protected] http://emergent.brynmawr.edu/mailman/listinfo/pyro-users
