Week 15 & 16 - Training CNN+LSTM for raw samples
The first approach to include recurrence in the prediction of modeled images is to use an LSTM layer after the CNN to analyze temporal correlations. The new defined network is as follows:
URM dataset results
I have trained and evaluated the network with the simplest set of all, URM, obtaining the following results
-Fix “y0”, 800 training samples, 100 test samples
When comparing these results with those obtained for CNN in this same dataset, in which no error was made (relative error = 0%), we realize that this strategy does not seem correct.