Mackey Glass time series prediction

You work on a certain topic or data set and don't know how to start off with it using MemBrain? Not sure if your net design matches your problem or if there is room for optimization? Is it reasonable at all to approach your problem with NNs? Is MemBrain the correct tool to accomplish your task and to match your infrastructure?

These questions are best placed here!
dinufira
Posts: 7
Joined: Thu 7. Oct 2010, 13:41

Re: Mackey Glass time series prediction

Post by dinufira » Fri 18. Feb 2011, 20:49

Hi, Thomas,

Thanks again for your patience. Ok understood, now all my time series are organised in .csv files as per your instructions (2 columns, the output one having the data that appears next). Now I have to clarify the following steps, in order to organise all files properly.
- as mentioned each time serie is organised as per your instructions, I left the first line unchanged (x(t);x(t+1)). I presume that this is the Train file for each time serie. If I am wrong please correct me.
- Now I have to prepare the validation data set. In your instructions it's mentioned that this has to be also with 2 columns but this time as I understand the first line is X(t);X(t+1). The question is how do I organise the data in the validation file ? I presume that even if has also 2 columns the data in the columns is not organised like in the Train file. You mention in your instructions : It is very important that the validation data makes up only a few data points and the majority of the data goes into the training lesson. Furthermore, the validation data points must take over seamlessly where the training data end, i.e. the validation represents the valid appendix of the training time series up to the last known data point in time. Please clarify.

I have to also clarify your instruction : Prepare at least one net with one input x(t) and one output x(t+1). I have the best trained net candidate - Net 5. Can I use this one ? What do you mean by at least one net ?

And one last thing to clarify for the moment, I quote : "After the script has executed, open the file 'Extrapolated.csv' in Excel, add a column to the left in order to have an X-Axis". Ok and then comes the part I have to understand, I quote again : "Then you can easily create a plot of the input and output data to the net over the whole time series, including the four extrapolated data points in the very end."

I think that if we clarify these 3 points I can organise the data properly and go on.
Awaiting with interest your answers, thanks again and best regards.

Dinu

User avatar
Admin
Site Admin
Posts: 438
Joined: Sun 16. Nov 2008, 18:21

Re: Mackey Glass time series prediction

Post by Admin » Wed 23. Feb 2011, 07:42

Hi Dinufira,
dinufira wrote: The question is how do I organise the data in the validation file ? I presume that even if has also 2 columns the data in the columns is not organised like in the Train file.
Yes, the data must be organized in exactly the same way as in the training file.
dinufira wrote:You mention in your instructions : It is very important that the validation data makes up only a few data points and the majority of the data goes into the training lesson. Furthermore, the validation data points must take over seamlessly where the training data end, i.e. the validation represents the valid appendix of the training time series up to the last known data point in time.
From my point of view the sentence is clear. Please ask more precisely what is not clear. If the whole (known) time series has 1000 data points, this makes up 999 lines of data organized in pairs of t; t+1 then put for example the first 900 data lines in the training file, the following 99 lines in the validation file.
dinufira wrote:I have to also clarify your instruction : Prepare at least one net with one input x(t) and one output x(t+1). I have the best trained net candidate - Net 5. Can I use this one ? What do you mean by at least one net ?
This is for the case if you wanted to design your own net candidates. You can, however, just use the existing candidates. Just leave them as they are. They will all be trained on your data through the script and the one that performs best on your data will be selected by the script automatically.
Admin wrote:Ok and then comes the part I have to understand, I quote again : "Then you can easily create a plot of the input and output data to the net over the whole time series, including the four extrapolated data points in the very end."
Again, what part of the sentence did you not understand? Did you try to perform this step? What went wrong?

Regards,

Post Reply