Neural Networks

Neural Networks from watertohoku on Vimeo.
Redes Neuronales
This presentation was prepared in Power Point, and recorded without audio. The video is at the top of this entry. Some questions asked during the session are also written below.

The acknowledgements go to (I am forced to reserve the name of my friend to avoid copyright issues).

Q&A


1. Artificial neural networks (ANN) & sensitivity analysis?
Sure, and not only for those kind of analysis, but also for any application where we desire to predictions a trend.

2. An aspect to be careful when applying ANN?
The selection of the inputs is critical. Thus, for the example of a model for the predictions of catchment discharge, we would not consider among the inputs variables unlikely to have some relationship (example: there is no relationship between the number of students working on the model and the total precipitation).

3. Having poor knowledge on the response of a given system, is there some function recommended to begin with the training process?
We may begin with the logistic curve. Various research papers recommend its use for initial attempts.

4. Is there some software we can use for teaching purposes?
Various. Nam san recommends to have a look on the "Tiberius". Simple enough for simple applications, simple enough to learn on the capabilities of the ANNs, and fully functional for a certain period for students enrolled to an academic institution. Tweet