Simulation and prediction of time series using convolutional neural networks

Hi everyone,
I'm working on the problem of downscaling precipitation from global circulation model (GCM) to observed local precipitation.
I'm asking how can CN be used for this purpse with:
The input data are time series of precipitation and temperature from GCM (size 1000x2)
The Target data is the local observed precipitation data (size 1000x1)
After training the network, I want to use this network to simulate the its reponse for new data inputs.

Answer 1

[Harsh]

Hi @Mahmoud Zemzami

To downscale precipitation using a cascade-forward neural network (CFNN) in MATLAB, start by formatting your GCM input data (precipitation and temperature) and observed local precipitation target data as time series, and apply appropriate preprocessing like normalization and data division. Refer to the following documentation:

Input/Output Processing Functions – https://www.mathworks.com/help/deeplearning/ug/choose-neural-network-input-output-processing-functions.html

Then, create and configure a CFNN using the built-in function described here:

cascadeforwardnet – https://www.mathworks.com/help/deeplearning/ref/cascadeforwardnet.html

After training the network, simulate responses for new input data using the standard training and simulation workflow. Refer to the following documentation:

train – https://www.mathworks.com/help/deeplearning/ref/train.html

sim – https://www.mathworks.com/help/deeplearning/ref/sim.html

I hope this resolves your query!