NN3 HomeResultsMotivationInstructionsDatasetsSubmissionFAQ & HelpConferencesPublications

FAQ

  • Do I have to attend a workshops or conference?
    No. You may also take part in the competition and only submit your predictions.You can still win!
  • Do I have to attend all workshops?
    No. The idea is to offer you choices when & where to present.
  • Can I win something?
    Yes ... fame through publications and by winning the competition or special awards, such as best student submission, best NN submission etc., but no fortune (yet)! We are talking to conference organisers to source funding for student submissions.
  • Are other methods allowed?
    Yes, as long as they are methods of computational Intelligence or computationally intensive (ridge regression, wavelets etc.) Statistical methods may submit results as benchmarks (as they are not CI methods) and will be evaluated against the computational methods! We welcome submissions from statistical software vendors to demonstrate the capabilities of their software, although they may not win the
    competition.
  • Do I have to forecast all series with the same Neural Network setup?
    No. The objective requires a single & consistent methodology, that is implemented across all time series. This does not require you to build a single neural network with a pre-specified input-, hidden and output-node structure but allows you to develop a process in which to run tests and determine a best setup for each time series. Hence you can come up with 111 different network architectures, fuzzy membership functions, mix of ensemble members etc. for your submission. However, the process should always lead to selecting the same final model structure as a rigorous process. So if you - for example - wish to differentiate between trended, seasonal and stationary time series in model building you must develop a methodology that systematically test each of the series in the same way to make successive decisions. You may of course also use just a single neural network architecture to forecast all time series identically. In other words, you must develop a methodology than can be replicated by others, and would lead to identical results on the same dataset.
  • ...

For additional help please contact the workshop organisers

Sven F. Crone
sven.crone@neural-forecasting.com

Lancaster University Management School
Centre for Forecasting
Lancaster
LA1 4YX

Tel +44.1524.5-92991
Fax +44.1524.xxxxxx

Thanks

We wish to express our gratitude to a number of reviewers for their helpful suggestions on the experimental design:

  • Robert Fildes, Lancaster University Management School
  • Attendees at the ISF'06 Neural Forecasting Track, Santander, Spain
  • Attendees at the ISF'05 Neural Forecasting Track, San Antonio, USA

We wish to thank the following for support in communication of the NN3 competition:

  • Ashu M.G. Solo, Maverick Technologies

 

©  2006 BI3S-lab - Hamburg, Germany - All rights reserved - Questions, Comments and Enquiries via eMail -  [Impressum & Disclaimer]
The Knowledge Portal on Forecasting with Neural Networks @ www.neural-forecasting.com -
last update: 18.10.2006