Time Series Data Encoding for Deep Learning, TensorFlow and Keras (10.1)

preview_player
Показать описание
Time series data is usually represented in the form of sequences when working with Keras and TensorFlow. In this video sequences are introduced for time series prediction.

Code for This Video:

Follow Me/Subscribe:

Рекомендации по теме
Комментарии
Автор

I have been binge watching your videos. They are simply brilliant. Thank you.

esra_erimez
Автор

I haven't done any serious time series modeling yet, just a few small projects using linear regression in scikit-learn. This video made me curious about the kind of performance you can get with LSTMs in TensorFlow. I'll give it a try. 😊 Thank you for another useful video.

BiancaAguglia
Автор

Correction: In each neuron u pass all the features, not every single feature is passed seperately in every single neuron.

smilebig
Автор

how would you encode a dataset containing multiple time series? like a series for temperature and one for light level.
I would like to use an autoencoder on this type of data.
For now, I am only concatenating all the values in a single using a min-max scaler on related columns.
If I plot the decoded time series they are not very similar to the input ones.

cunningham.s_law
Автор

"encoding" not a great title choice. more like "shaping"

LayneSadler
Автор

Can you suggest any method I could use to machine learn a 2D time series output ?

I am trying to learn the movement of a biofilament based on its Stiffness/velocity/ amplitude.

I have a model that can generate this data, I was hoping to learn to avoid the slow model.

Is it possible to ml

Input (amplitue/stiffness/time) ---> (x-y) movement in time

snaidu
Автор

Hello, thanks for all but I have a request please use real data instead of creating vectors like X you start by two value and after that, you go back to one and you don't explain well how reshaping data to get the right dimension.
Thanks again.

maloukemallouke
visit shbcf.ru