Informer: Complete Code from scratch!

We are going to take a look at the time series transformer line by line

ABOUT ME

RESOURCES

PLAYLISTS FROM MY CHANNEL

MATH COURSES (7 day free trial)

OTHER RELATED COURSES (7 day free trial)

[sudlow]

Really clear and helpful, thanks. I'm a bit hazy on Einstein summation but I got the general idea!

[小飞侠-de]

Amazing video, you are a genius, please keep releasing new videos. you are the most handsome and smartest person in the machine learning field.

[tomoki-vo]

I was very slow at thinking and when using normal calculator and at exams, so i bought a programmable calculator Casio . the problem that i always needed to debug the programmes that i prepared during the exams (due to minor unexpected changes in the exams questions ) .so i get medium grades after all.
Thats how i learned programming in secondary school . the language i used BASIC.

[scott]

Amazing video. I'll try run this model on my comp and then adapt it to my project. My data is quarterly though so freq='qtr' I also want to forecast 2 years ahead so out_len = 8. I'll set seq_len to the number of observations or rows in my training data which is around 128. Why is kernel_size = 3 wouldn't you want your model to look back as far as possible when making a prediction? I'm going to set mine to at least 8. Hopefully I can update all those parameters and the model still runs. I'll post again later to let you know how I get on :)

[scott]

There is no quarterly option in the colab notes: # freq for time features encoding, options:[s:secondly, t:minutely, h:hourly, d:daily, b:business days, w:weekly, m:monthly]. How do I go about getting quarterly freq?