The sampling frequency of multiple data is different, how to process the data is better? For example, the following picture
First thing I’d consider is aggregation, so for instance take the mean of every 10 time steps (if it’s likely that there will be at least a couple real values every 10).
This will obviously reduce the total number of data points, but achieve the purpose of making every data point represent the same amount of time. This should generally help the signal to noise ratio, which helps the TM learn the sequences with less data.