I'm looking to train one self-organizing map on different trajectories which are basically data frames of different lengths. It works very well with one trajectory but it leads to massive overfitting and doesn't help to recognize different types of trajectories. My aim is for the SOM to classify the points in the trajectories as different types of STOPS & MOVES.
I tried putting the trajectories in lists but I get an error as the objects are not of similar length.
Hope you can help :),
Baptiste
