[pymvpa] Fw: Hyperalignment
Salim Al-wasity
salim_alwasity at yahoo.com
Thu Dec 1 16:46:28 UTC 2016
Hi SwaroopI checked the commonspace in both cases and it turns out that they are not identical, but almost equal as shown below. Therefore, the realigned task datasets (ds_task) are totally different in both scenarios.
>>> np.allcolse(hyper1.commonspace, hyper2.commonspace, rtol=1e-3)>>> True
>>> np.allcolse(hyper1.commonspace, hyper2.commonspace, rtol=1e-4)>>> False
>>> err=np.isclose(hyper1.commonspace, hyper2.commonspace, rtol=1e-4)>>> err.size-np.sum(err)>>> 8>>> err=np.isclose(hyper1.commonspace, hyper2.commonspace, rtol=1e-5)>>> err.size-np.sum(err)>>> 141>>> err=np.isclose(hyper1.commonspace, hyper2.commonspace, rtol=1e-6)>>> err.size-np.sum(err)>>> 1613
On Thursday, 1 December 2016, 2:37, Swaroop Guntupalli <swaroopgj at gmail.com> wrote:
Ok. Can you check if data in hyper.commonspace is the same after training in the two scenarios?
On Wed, Nov 30, 2016 at 8:08 PM, Salim Al-wasity <salim_alwasity at yahoo.com> wrote:
Hi SwaroopDo you mean like this:
Scenario-3-:new_ds_task=[ds_task[2], ds_task[0], ds_task[1], ds_task[3], ds_task[4], ds_task[5], ds_task[6], ds_task[7], ds_task[8], ds_task[9]]
new_ds_h=[ds_h[2], ds_h[0], ds_h[1], ds_h[3], ds_h[4], ds_h[5], ds_h[6], ds_h[7], ds_h[8], ds_h[9]]
hyper=Hyperalignment()
hypermaps=hyper(new_ds_h)ds_hyper = [ hypmaps[i].forward(d_all) for i, d_all in enumerate(new_ds_task)]
Same result as scenario 2, still different than scenario 1!.
Sent from Yahoo Mail on Android
On Thu, 1 Dec, 2016 at 12:55 am, Swaroop Guntupalli<swaroopgj at gmail.com> wrote:
I had 10 subjects whose Hyperalignment data are stored in (ds_h) and task data in (ds_task) which I am classifying
Scenario-1-:
hyper=Hyperalignment(ref_ds=2)hypermaps=hyper(ds_h)ds_hyper = [ hypmaps[i].forward(d_all) for i, d_all in enumerate(ds_task)]..........Then continue with classification
Scenario-2-:new_ds_h=[ds_h[2], ds_h[0], ds_h[1], ds_h[3], ds_h[4], ds_h[5], ds_h[6], ds_h[7], ds_h[8], ds_h[9]]hyper=Hyperalignment()hyper.train(new_ds_h)
hypermaps=hyper(ds_h)
Can you combine the above lines into one call and re-order the mappers and see how it behaves?
ds_hyper = [ hypmaps[i].forward(d_all) for i, d_all in enumerate(ds_task)]..........Then continue with classification
SincerelySalim
______________________________ _________________
Pkg-ExpPsy-PyMVPA mailing list
Pkg-ExpPsy-PyMVPA at lists. alioth.debian.org
http://lists.alioth.debian. org/cgi-bin/mailman/listinfo/ pkg-exppsy-pymvpa
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20161201/d1f5a450/attachment.html>
More information about the Pkg-ExpPsy-PyMVPA
mailing list