Neural network alternatives

[afr2903]

Given time constrain, a static neural network will be used. However, after discussing with the mentor, alternatives were presented:

• Feed-forward neural network: Good for time-series as takes into account previous steps, sample time should be constant.
• Recurrent neural networks: Better if sample time is noisy.
• Informed neural networks: Follow a physical equation to take load off the non-linear data.

Will be explored later for the research paper.

[afr2903]

Time series NN

Feedforward tutorial