Location:
Darwin 102

Agenda:
NEW TRAINING METHODS FOR NEURAL NETWORKS
Neural networks attempt to learn solutions to problems by emulating
simplified neurons. They are becoming increasingly popular for
helping solve complex problems such as vehicle guidance, face
identification, handwritten text recognition, and spam filtering.
This talk presents research into new neural network training methods
such as particle swarm optimization (PSO) and neighbor annealing,
combined with a divide-and-conquer approach, that may enable neural
networks to solve even larger problems.