It'd be great to have a few examples of taking a relatively hard task for deep nets to learn, implementing it in Nengo, and then using those connection weights as your starting point for deep learning.
I'm not sure what tasks, but I'm thinking something with recurrence. I saw one poster where people were feeding sequential inputs like:
( ( 3 + 4 ) - ( 7 - 2 ) )
and expecting an output of 2
That task might work, but I'm not sure it's the best example.
It'd be great to have a few examples of taking a relatively hard task for deep nets to learn, implementing it in Nengo, and then using those connection weights as your starting point for deep learning.
I'm not sure what tasks, but I'm thinking something with recurrence. I saw one poster where people were feeding sequential inputs like:
and expecting an output of
2That task might work, but I'm not sure it's the best example.