This repository is NOT up-to-date. Use under your own risk :-)

Este es un repositorio con código de ejemplos sobre keras o tensorflow para LSTM, GRU, RNN.

• LSTM en KERAS https://keras.io/getting-started/sequential-model-guide/
• entrada o input multivariable http://stackoverflow.com/questions/40331510/how-to-stack-multiple-lstm-in-keras
• Learn Markdown
• recurrent layer in keras https://keras.io/layers/recurrent/
• channels dimension for timeseries

• boston housing regression example https://github.com/benbusse/keras-housing-regression/blob/master/keras-housing-regression.ipynb
• keras intel desafio kaggle para clasificar en tres tipos de cervix, imagenes https://www.kaggle.com/pierretisseur/intel-mobileodt-cervical-cancer-screening/keras-intel/code
• Ejemplos tomados de este enlace http://machinelearningmastery.com/time-series-prediction-lstm-recurrent-neural-networks-python-keras/
• Time series prediction with multiple sequence input https://groups.google.com/forum/#!searchin/keras-users/time$20series|sort:relevance/keras-users/9GsDwkSdqBg/2BVSits77yEJ

input_shape=(1, 1)

The first number is the number of timesteps the model should expect. Given only one timestep of data input the best possible approximation of any moving average is just to repeat the unique input value.

The second number is the number of inputs the model should expect at each timestep. This seems ok given your stated goal.

Your input has to be a list where each array has the batch size as the first dimension. Hence:

[(batch_size, 45, 8, 3),(batch_size, 45, 8, 3)...]

Whether or not you should use an Activation as the last layer, and what kind of activation, depends on the range of the values you want to output (for instance: if you want to output negative and positive values, don't use ReLU, etc. And never use softmax since it ouputs a probability distribution).

If you aren't sure, it's probably better not to use an Activation as the last layer (Dense would then be the last layer).

Also, "show_accuracy" should not be set for a regression problem. The notion of accuracy only makes sense for a classification problem.