Testing error in demo_1_xor.py really large

[indraneelpatil]

Hello!

Thanks for your implementation of the Neural Network, it is really helpful in my project.

I tested the demo you have described in the ReadME file of the XOR dataset but after running the demo_1_xor.py file the network is not able to predict the targets of the test dataset at all as seen here:

Testing...
+------------------------------------------------------------+
[#0] (1, 0) -> [0.4983276267402251] (targets=[1])
[#1] (0, 0) -> [0.5061470781947205] (targets=[0])

Any idea what is wrong here? From the learning curves it seems like the network is learning properly but these predictions are pretty off. Do you think the architecture of the network needs to be more complex for XOR? or is there any other problem?

Thanks,
Indraneel.