Input weights shape

[filippomb]

Input weights shape

[FilippoMB]

I'm not sure if I understand your problem.
'V' is the size of the input, i.e., the number of input variables.
'internal_units' denotes the size of the reservoir, i.e., the number of output variables.
Therefore, the weights must have shape (V, internal_units), as they map the input from a space of dimensionality 'V' to a space of dimensionality 'internal_units'.

'N' is the number of samples. In reservoir computing (and generally in neural nets) the weights depend on the dimensionality of the samples, not on the number of samples.

[NorhenAbdennadher]

Thank you for your answer, I think in this case, I'm finding troubles understanding the difference between N, T and V in the MTS. I'll be grateful if you can provide me explanation about the difference between T and V especially.

…

On Fri, 23 Jul 2021, 10:16 Filippo Bianchi, ***@***.***> wrote: I'm not sure if I understand your problem. 'V' is the size of the input, i.e., the number of input variables. 'internal_units' denotes the size of the reservoir, i.e., the number of output variables. Therefore, the weights must have shape (V, internal_units), as they map the input from a space of dimensionality 'V' to a space of dimensionality 'internal_units'. 'N' is the number of samples. In reservoir computing (and generally in neural nets) the weights depend on the dimensionality of the samples, not on the number of samples. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#7 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ATHFKYJ5TPHO7VOQB3E4P5LTZEXPJANCNFSM5A24DZXQ> .

[FilippoMB]

• N is the number of the time series in the dataset.
• T is the length of the time series, i.e., the number of time steps.
• V is the number of variables, i.e., how many quantities are measured at each time step. For example, if the time series describes a trajectory of an object in a 3D space, V=3 because you are measuring 3 coordinates at each time step. If the time series is univariate, then V=1.

You can read more about multivariate time series online. There are many tutorials and explanations.