The problem with neural networks
This, from the always interesting Jack Crossfire blog, is the smartest observation on neural networks and genetic algorithms I've seen in ages. Excerpt:
"In our last experience with neural network feedback, the feedback always headed towards 0. That one used straight back propagation to predict cyclic from attitude change. That fiasco has cursed all future investments in the radically different genetic approach. The genetic approach would constantly evaluate the attitude tracking & standardize networks which track better. How do U keep it from forgetting high wind feedback when evolving in low wind?"
To repeat: How do you keep it from forgetting high-wind feedback when evolving in low wind?
As humans, we solve this with a mix of short term and long term memory. Do we need to create an analog in neural computing to avoid the problem Jack rightly spots?
"In our last experience with neural network feedback, the feedback always headed towards 0. That one used straight back propagation to predict cyclic from attitude change. That fiasco has cursed all future investments in the radically different genetic approach. The genetic approach would constantly evaluate the attitude tracking & standardize networks which track better. How do U keep it from forgetting high wind feedback when evolving in low wind?"
To repeat: How do you keep it from forgetting high-wind feedback when evolving in low wind?
As humans, we solve this with a mix of short term and long term memory. Do we need to create an analog in neural computing to avoid the problem Jack rightly spots?
-
Comment by Howard Gordon on June 25, 2008 at 12:33pm -
You have to bear in mind that a neural network is just a statistical method that attempts to create a useful mapping between sensor input and control output. Its intelligence derives from the manner by which it is trained, and training data sets need to include old experiences as well as new. Otherwise, the network will tend to unlearn old lessons, which is an issue found with recurrent neural nets (a net that adapts in real time).
-
3D Robotics Comment by Chris Anderson on June 25, 2008 at 1:19pm -
That suggests that the idea of purely real-time learning neural network isn't practical. If you have to combine real-time learning with replaying old data sets (hmmm: real life vs dreams? Sorry to anthropomorphize!) it sounds like more trouble than it's worth.
-
Comment by Howard Gordon on June 25, 2008 at 1:53pm
-
I work with fairly sparse training sets, so I think it would not be too difficult to balance old and new. I would have to experiment with this idea.
Comment
Social Networking
Top Content
Contests
Season Two of the Trust Time Trial (T3) Contest has now begun. The fourth round is an accuracy round for multicopters, which requires contestants to fly a cube. The deadline is April 14th.A list of all T3 contests is here.
Groups
-
Melbourne
48 members
-
Documentation Editors
51 members
-
ArduCopter User Group
1299 members
-
Moderators
24 members
-
UK Ardupilot Group
111 members
Advertisement
© 2013 Created by Chris Anderson.
Powered by
You need to be a member of DIY Drones to add comments!
Join DIY Drones