Week 18 & 19 - Training ConvLSTM for raw samples

less than 1 minute read

Proposed net

Continuing with the use of raw images recurrence propose the use of ConvLSTM layer resulting in the following net:

ConvLSTM Net

Before the recurrent layer, it is necessary to introduce a convolve to reduce the dimensionality of the data and make it more manageable.

Linear dataset results

  • 1 DOF, 8000 training samples, 1000 test samples
ConvLSTM-Linear 1 DOF
  • 2 DOF, 8000 training samples, 10000 test samples
ConvLSTM-Linear 2 DOF

Parabolic dataset results

  • 1 DOF, 8000 training samples, 10000 test samples
ConvLSTM-Parabolic 1 DOF
  • 2 DOF, 8000 training samples, 10000 test samples
ConvLSTM-Parabolic 2 DOF
  • 3 DOF, 80000 training samples, 10000 test samples
ConvLSTM-Parabolic 3 DOF

Sinusoidal dataset results

  • 1 DOF, 80000 training samples, 10000 test samples
ConvLSTM-Sinusoidal 1 DOF
  • 2 DOF, 80000 training samples, 10000 test samples
ConvLSTM-Sinusoidal 2 DOF
  • 3 DOF, 80000 training samples, 10000 test samples
ConvLSTM-Sinusoidal 3 DOF
  • 4 DOF, 80000 training samples, 10000 test samples
ConvLSTM-Sinusoidal 4 DOF

Conclusions

The results obtained are better than those of the CNN, although there is still a prediction limit in the prabolic 3 DOF and in the sinusoidal 2 DOF, although with a smaller error. It should also be noted that, although low average values are obtained, the maximums are high and there is a large presence of outliers that worsen the prediction.