I am only on Lesson 3, so this might be premature, but I have an interest in comparing time series data tables/charts, looking for the output to be - ‘hey , these are really similar (not identical), if you moved one of these x click to the left or right, they would correlate well’. Obviously, an application of this would be to find one chart that has predictive value to another chart in time. Not sure how to use Deep Learning to this. Any thoughts?

Hi Steven. While I am by no means an expert in this domain, I can tell you that this question has been thoroughly investigated long before deep learning was invented. You might check out “time series”, “forecasting”, and “cross correlation”.

As for applying deep learning, a good resource is the Time Series Study Group on these forums. DL models may well be able to find factors that are not easily visible otherwise. But for finding shift correlations, IMO the well-researched existing techniques will be faster and more reliable.

HTH, Malcolm

Thank you. I have spent about a decade researching the non-ML techniques, but it is well trodden ground. So indeed I am looking for the invisible factors, for which ML may be a good tool. But thanks for the the Time Series Study Group tip.