Seminar: Competitive Cooperative Neuro-Evolution (…) for Climate Change and Energy - Discipline of Computer Science and Information Systems

Home / Research Seminar / Seminar: Competitive Cooperative Neuro-Evolution (…) for Climate Change and Energy

The next research seminar takes place on 28/04/15, at the ICT conference room, from 16-17, as usual (light refreshment served at 15:45).

Speaker: Dr. Rohitash Chandra

Title: “Competitive Cooperative Neuro-Evolution with Applications in Time Series Prediction for Climate Change and Energy”

Abstract:

Cooperative co-evolution is a nature-inspired optimization method that divides a problem into sub-components and been used for neuro-evolution. It employs different problem decomposition methods to decompose the neural network training problem into sub-components.

Different problem decomposition methods have features that are helpful at different stages in the evolutionary process. A novel island based competitive cooperative co-evolution framework was proposed in our earlier research that takes advantage of competitive and collaboration in the neuro-evolutionary process.

Cooperative-neuroevolutionary methods have been applied for real world problems such as time series prediction for energy load and climate change problems such as cyclones. We have used co-evolutionary recurrent neural networks for cyclone track, wind intensity and combined track and wind intensity prediction. We have also used the same method for energy load forecasting. The seminar will give highlights of several papers that have been published recently and outline of the road ahead in using co-evolutionary neural systems to solve difficult problems.