1
$\begingroup$

There are some time series data like this: 

Time    Feature A     Feature B       Class
1          0.3           0.6            1
2          0.5           -0.2           0
3          0.8           0.3            1
...

When using RNN on MNIST, we can use 28 time steps of 28 rows as input. For these time series data, one could use 10 time steps of rows for the classification problem. (eg. Feed the rows from T1 to T10 one by one into the RNN to predict the class at T11, then the rows from T2 to T11 to predict T12 and so on.)

What is some way to convert the time series data to build an array of arrays?

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ Hey, Hendo welcome to Stackoverflow, I read through your question but was unable to understand your question, Could you please add some description probably simple to explain your question? $\endgroup$
    – Kaustubh
    Commented Sep 17, 2018 at 6:03
  • $\begingroup$ Indeed, the MNIST data set has no temporal element nor natural sequence. It's not clear what the goal is as MNIST is already naturally represented as 2d arrays. $\endgroup$
    – Sean Owen
    Commented Sep 19, 2018 at 15:40
  • $\begingroup$ Wrong, i any is extremely clear. Just you wrongx, can't understand, idts. $\endgroup$
    – frt132
    Commented Oct 2, 2018 at 9:13

0

Reset to default

Browse other questions tagged or ask your own question.