Transcript Group Learning vs. Individual Learning

Group Learning
Neural Networks: sharing information
By Philip Sterne
Supervisor : Shaun Bangay
Group Learning

Create virtual Creatures™.
 Give them the ability to learn
through Neural Networks.
 Place them in Stressful
environments.
 Group Learning vs.
Individual Learning
Artificial Life:

Moves around and explores its
environment.
 Communicates with other
creatures.
 Has needs,
eg – Hunger, thirst

Dies if those needs are not met.
Neural Networks

Attempt to model the brain.
 Just like the brain:
– Need to be trained.
– Can learn new things.
– Can forget old things.

Perform parallel computations.
Neural Networks
Jargon:
 Inputs (perceptions)
 Perceptrons
 Neurons
 Pathways
 Outputs
Stressful Environments

Predators
– Detection.
– Evasion.
– Warn others. (?)
– Remove slow learners.

Scarce Food
– Forces exploration.
– Encourages Cooperation (?)
Group vs. Individual Learning

Heart of my Project.
 In real life:
– People learn from other people.
– “Birds of a feather flock together”

In Neural nets:
– “Bad” nets are replaced with “Good” nets, but
only after death.
– No real interaction.
Group Learning

Enable communication.
– Pass knowledge (weightings)
– Speak to each other.

Ensure diversity.
– Only “like-minded” nets sharing weightings.
At Present:

Neural networks simulate individual
learning.
 No research (yet) on cooperative learning.
 Focused on
– more accurate simulation of the brain. E.g. –
Biochemical.
– Genetic encoding.
Comparisons:

Speed of Learning.
 Population Dynamics.
 Cooperation.
 Communication.
 End result.
Conclusion:

Create Neural Networks.
 Simulate social learning
 Will Language develop?
 Will there be cooperation?
 Will social learning be faster than in a group
of individuals?