Influence of the individual's size on the island model architecture

Abstrakt

Our motivation starts from recommendation systems on large data. We use hybrid evolutionary agent based system.

Our system runs in parallel several optimization methods. Main challenge is the frequency of individuals migration.

We describe technical restrictions which complicate parallelization techniques of helping each other computational methods. Size of data influences latency in communication and measurement.

We experiment with architectural changes which enable to move the limits. At the same time we would like to give planners sufficient freedom to manage distributed computation.

Klíčová slova

• Recommendation systems
• large data
• evolution based island models
• computation parallelization
• agent based systems.