Questions Regarding Dataset

[Leonardo Alchieri]

Hi everyone,

First off, I would like to thank the organizers for providing us with good data and an interesting challenge. 

I would like to pose a few questions if possible regarding the data. If the questions are already answered on the website, please accept my sincere apologies. 

1. For both the "Deep Features" and the "Hand-Crafted Features", there are some arrays called `masking`, e.g. `'ECG_masking'`. Would it be possible to let us know what does the masking refers to?

2. The Challenge Website, under the description of the Dataset, lists 8 GSR features (Table 1). However, in the actual data the dimensionality of the GSR features is 12, I am not sure what the remaining 4 feature mentioned in the website refer to?

3. I was wondering if there is some information, or if it could be made available, regarding which hand-crafted features are which. More specifically, what is the mapping between the description on the website (Table 1) and the data? Are they in the same order, e.g. the 1st column of the ECG features is the #1 feature in Table 1?

4. I was wondering if there is some information, or if it could be made available, regarding which hand-crafted features belong to which participant (anonymously).

I hope my questions are clear and understandable. 

I am sorry for the multiple questions. Thank you very much and

Best regards,

Leonardo Alchieri

PhD Student, Università della Svizzera Italiana, Switzerland

[Maryam Khalid]

1. The masks are used when deep features are  available for some data points. For example, if at 13:00:00, there is no qualified ECG data for us to extract the deep feature, then we will have zeroes in deep features, whereas the mask will be marked as 0. In short, a positive mask means the deep features are available at this timestamp.

2&3. Table 1 is modified on our website as we also included the skin temperature features along with GSR. The order of ECG features in the released data is the same as we showed on website. For the released GSR features, dimension 1-8 correspond to the 8 GSR features, and 9-12 are the 4 ST features in table 1.

4. We are still discussing on releasing participants ids. The current setting of the challenge is participant-independent.

[Tushar Agarwal]

Hi Maryam,
Will it be possible for you guys to release ECG data corresponding to these features?

[Leonardo Alchieri]

Hi Maryam,

Thank you very much for your kind and thorough reply. 

Best regards,

Leonardo

[Compwell Group]

Hello Leonardo,

Thanks for bringing this to attention.

Unfortunately, we can not share raw ECG data for this challenge.

Let us know if you have other questions or concerns.

Best regards,

Maryam

[Zac Dair]

Hi Maryam,

Can I confirm that the stress labels are one per hour? With essentially 60 sets of extracted features over the hour and one label for all of those?
Any insight into this would be greatly appreciated.

Kind regards,

Zac

[Leonardo Alchieri]

Hi Maryam,

Thank you very much for your reply.

All the best,

Leonardo

[Leonardo Alchieri]

Hi everyone,

Sorry to bother, but since working with the data I have chanced myself upon some other questions: I would be immensly grateful if someone might be able to answer them.

The data, for all of the features, is in the range [0,1], which I guess means it has been normalized: is this the case? If yes, do you know what kind of normalization was performed? Again, if this is mentioned somewhere, please accept my apologies.
I have found that, in some features, the 60 1-min long steps are all 0s. While it does not pose any kind of problem, I was just wondering what the reason behind it could be.

Thank you again for the helpfullness and I wish you my very
Best Regards,
Leonardo