How do I train multiple timeseries at the same time?

I have several different time-series of different shapes(same number of samples), how would I train a model using several of these?

For example, say

>>>Series_1.shape
(29276, 32, 4)
>>>Series_2.shape
(29276, 5, 14)

The first layer in the model would look like:

self.CNN = tf.keras.Sequential([
            layers.Conv1D(filters=32, kernel_size=3, activation="relu", input_shape=(features, steps), padding='same'),
            layers.MaxPool1D(pool_size=3, padding='same'),
# ... etc

However, the input_shape would only work on one input. Am I supposed to use a 2D Convolution for this task?

Hi @itaymr

Welcome to the TensorFlow Forum!

You may need to apply different techniques to train a model for multiple time-series with different shapes. You can do this either by reshaping the timeseries to have the same number of dimesions using dimensionality reduction algorithms or padding/slicing them to a common length and then concatenate the reshaped series to feed into the model training OR can train a different model on each individual time-series.