Time series/ sequential data study group

Part 1 (2019)
2

To kick off this thread, I’d like to send you a link to a post I wrote in the ‘Share your work’ thread (Time series classification: Transfer Learning with Convolutional Neural Networks). Apologies if you have already seen it!

It shows one way in which a univariate TS dataset (OliveOil) -from the UCR time series datasets- can be transformed into images (using Gramian Angular Field), and then modelled following the general transfer learning approach we used in lessons 1-2.

The results surprised me (very close to state of the art!) considering:

  1. How small the train sample is (30 samples only)
  2. These GAFD images are very different from those in Imagenet.
  3. I was just applying the standard fastai method, only tuning epochs and lr. I’m sure there is room for improvement.
43 Likes