TimeSeries

fastai fastai dev
130

Thanks for sharing @kobus78.
Your notebook is a great example of how time series to image transforms can really add value!!!
This really motivates me to keep improving and adding new functionality to the tsai library.

3 Likes
131

wow, good work! this could be a paper!
I have a couple of questions:
Did you tried Rocket? or something simpler like a 2channel CNN as baseline?
I don’t know the resolution you have on your time axis, but transforming to images can make you loose a lot of info due to pixel resolution limitations.
When you said inference, do you used some pytorch backend for serving your models? (torchscript, tensor RT, onnx, etc…?) I am curious how those work for TS.

2 Likes
132

Hi Thomas,

Thanks for the feedback.

I did not try Rocket or something simpler - I specifically wanted to use the opportunity to dig into fastai.

Thanks for the comment about losing resolution - something to keep in mind. In my case, though, airliners change altitude in significant steps - the only changes I were interested in for this problem. Also, a step change takes substantial time relative to a pixel’s resolution.

By inference, I simply meant subjecting the models to unseen test images, i.e. I did not deploy them to production.

  • Kobus
1 Like
133

Thanks for your reply @oguiza. I actually forgot to add another example: This one also uses the principle of first converting a time-series to an image, but, instead of performing classification, it performs regression. Please see:

Time-series Regression (Deep Learning to Detect Change Points)

Granted, this approach is probably an overkill. Also, it is probably not as accurate as more traditional approaches. But, I was curious about how it would do. This report links to the actual notebook. At the bottom of the notebook I have visualized the regression predictions in green dotted lines: notebook

  • Kobus
3 Likes
134

A faster Rocket model is avialable:

8 minutes to run the full UCR dataset!!!

2 Likes
135

Not presupposing any particular interest in this, but as I said here, I’m happy to answer any questions anyone might have, help with troubleshooting if you want to try it out, and also just hear any feedback.

1 Like
136

Synthetic Time-Series Data: A GAN approach

https://towardsdatascience.com/synthetic-time-series-data-a-gan-approach-869a984f2239

2 Likes
137

Meeting with Angus Dempster - Rocket, MiniRocket, and MultiRocket

Hi all,

I’d like to invite you to participate in a web meeting we’ll have with Angus Dempster next week (@angusde ).

For those of you who don’t know him, Angus is a Ph.D. student at Monash University in Australia (a world-class group in time series research) and is one of the authors of several outstanding papers in Time Series Classification.

If interested, please read this.

138

@oguiza Does your FastAI based TS library include “tft” implementation, like the one in pytorch ? Do you have a separate dataset/dataloader to handle it ?

139

Not yet, @vrodriguezf is working on t!t

140

I was advised by @tcapelle to share my Pytorch based probabilistic time series framework work here. Kindly have a look at https://github.com/zalandoresearch/pytorch-ts for some state of the art uni and multivariate time series methods that are implemented. Thanks!

6 Likes
141

Welcome, take a tour also here. You will find a lot of resources.

1 Like
143

thank you @kashif!!

1 Like
144

Hi @oguiza, hope you are fine.
Can you suggest me some work for stock-price prediction(or crypto-price prediction).Can it be done using tsai?
And, most importantly, do these models(i.e if someone is really benefitting from it), or it is nearly impossible to make accurate stock-price prediction models?

Thanks…

145

Hi, hope all of you are fine.
Kindly help me with making the TimeSeries Dataset for Crypto Prices Data(including Open,High,Low,Close,Volume,MarketCap).

Thanks…

146

Is there any plan to include MultiRocket implementation in tsai?
It is the second best SOTA currently after HIVE.

I can try implementing it. Maybe starting from MiniRocket implementation? Any tips or repos to share anyone?

147

That’d be great. MiniRocket is a good starting point. You may also take a look at MiniRocketFeaturesPlus and MiniRocketPlus. One of the arguments available is add_lsaz (longest subsequence above zero). This is the feature that scored best in MultiRocket. But there are many other options that can be used with MultiRocket (which makes it more complex too).
Please, feel free to submit a PR with a MultiRocket implementation if you decide to move forward.

148

Can you tell me how MiniRocket is in sklearn while the library itself is on PyTorch

149

First of all, great implementation @oguiza, and credits to the people who have contributed to the project. I appreciated the effort.

I was wondering about the idea behind using None in item transformations such as [None, [TSRegression()]] or [None, [Categorize()]]. I can’t truly remember but I might also have seen without None versions of them.

150

The first tfm in tfms is applied to the X and the 2nd to the y. X usually doesn’t require a transform. That’s why it’s usually None.

1 Like