<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"><html><head><meta content="text/html;charset=UTF-8" http-equiv="Content-Type"></head><body ><div style='font-size:10pt;font-family:Verdana,Arial,Helvetica,sans-serif;'><div>Thanks for the suggestion, Benoit! I think the easiest way to achieve that is sample a quaternion uniformly on a unit 4D sphere. <br></div><div><br></div><div>By the way I updated the ideas wiki and I will include your suggestion as well. Also as was suggested we could use a more general name scipy.spatial.transform.<br></div><div><br></div><div>Nikolay</div><div class="zmail_extra"><div id="Zm-_Id_-Sgn1"><div><br></div><div><br></div></div><blockquote style="border-left: 1px solid #cccccc; padding-left: 6px; margin:0 0 0 5px"><div><div>Hi,<br></div><div> <br></div><div> I have used that library quite a few times, and it is rather slow.     Adding a transformation (or, for starters, rotation) module to scipy     would be, in my opinion, a nice addition. <br></div><div> <br></div><div> Speaking about adding a few algorithms, one interesting add could be     a function to uniformly sample the rotation space. It is a core     functionality needed in a number of cases, and not that     straightforward to perform properly (again, depending on the chosen     formalism for representing the rotation). <br></div><div> <br></div><div> Best<br></div><div> Benoit<br></div><div> <br></div><div> <br></div><div> <br></div><div class="moz-cite-prefix">On 19/01/2018 22:28, oss wrote:<br></div><blockquote><div>Hello, <br></div><div class=""><br></div><div class="">Maybe this could come in handy regarding transforms         matrices quaternions<br></div><div class=""> etc.<br></div><div class=""><br></div><div class=""><a class="" href="https://www.lfd.uci.edu/%7Egohlke/code/transformations.py.html">https://www.lfd.uci.edu/~gohlke/code/transformations.py.html</a><br></div><div class=""><br></div><div class="">Best<br></div><div class=""><br></div><div class="">Tryfon<br></div><div class=""><div><br></div><div><div><br></div><blockquote class=""><div class="">On Jan 19, 2018, at 4:25 PM, Eric Larson <<a class="" href="mailto:larson.eric.d@gmail.com">larson.eric.d@gmail.com</a>>               wrote:<br></div><div><br></div><div class=""><div class="" dir="ltr"><div>I have personally run into the                 need for such transformations in two separate domains                 (3D visualization, neuroscience / electrophysiology) and                 I know it's used in multiple other places, too. So I                 think it would be sufficiently general. I'd look forward                 to having it in SciPy! <br></div><div class=""><br></div><div class="">I'd be happy to be a secondary mentor on                   this if you (or someone else) wants to be primary.<br></div><div class=""><br></div><div class=""><div>Best, <br></div><div class="">Eric<br></div><div class=""><br></div></div></div><div class="gmail_extra"><div><br></div><div class="gmail_quote"><div>On Fri, Jan 19, 2018 at 10:11                   AM, Nikolay Mayorov <span class=""><<a class="" href="mailto:nikolay.mayorov@zoho.com">nikolay.mayorov@zoho.com</a>></span> wrote:<br></div><div> <br></div><blockquote style="margin: 0 0 0 0.8ex; border-left: 1px rgb(204, 204, 204) solid; padding-left: 1ex" class="gmail_quote"><div class=""><div class="" style="font-size: 10pt; font-family: Verdana, Arial, Helvetica, sans-serif"><div class=""><div class=""><div>Hi!<br></div><div> <br></div><div> I have this idea, which I'm well familiar                             with. The module would be called like                             scipy.spatial.rotation and be devoted to the                             rotation formalism in 3 dimensions. <br></div><div> <br></div><div> The main objects are Euler angles (and their                             variations), direction cosine matrices,                             quaternions and rotation vectors. We can go                             with an abstraction class Rotation (using                             DCMs or Qs internally), but we should be                             able to create that from any representation                             and see it in any representation. In spirit                             of scipy/numpy we use vectorized/bulk                             approaches (i.e. many rotations in single                             Rotation class).<br></div><div> <br></div><div> Rotation should support 2 operations:                             compose 2 consecutive rotations and                             rotate/project a 3d vector.  Of course all                             procedures must be 100% robust and there are                             some fine points, especially in conversions                             between representations.Also we can add some                             algorithms, like: quaternion interpolation                             (SLERP), least-squares vector matching by a                             rotation (Whabba's problem), more advanced                             and less known algotithms for rotation                             interpolation, and I will try to come up                             with something more. <br></div><div class=""><br></div><div class="">Overall it seems reasonably                               straightforward , but with enough                               challenges in design and implementation<br></div><div class=""><br></div><div class="">As currently described, it                               might be not enough volume for the GSoC,                               but we can develop it farther. <br></div><div class=""><br></div><div class=""><div>Also I'm not sure if its                               applicability is broad enough to include                               it into scipy. I believe similar                               functionality is available in Aerospace                               toolbox in Matlab. I want to hear some                               opinions on that.<br></div><div> <br></div><div class=""><br></div><div class="">Nikolay<br></div><div class=""><br></div><div class=""><div class=""><div><br></div><div>---- On Wed, 10 Jan 2018 17:02:33                                   +0500<b class=""> <a class="" href="mailto:jomsdev@gmail.com">jomsdev@gmail.com</a> </b> wrote ----<br></div><span class=""><div><br></div><div><br></div><blockquote class="" style="border-left: 1px solid rgb(204, 204, 204); padding-left: 6px; margin-left: 5px"><div class=""><div class="" dir="ltr"><div class=""><div class=""><div class=""><div class=""><div>Hi all,<br></div><div> <br></div></div><div>Last year I started                                                 implementing some                                                 methods for Randomized                                                 Numerical Linear Algebra                                                 (RNLA) in scipy.<br></div><div> <br></div></div><div>By now it is only the                                               CountMin Sketch                                               (clarkson_woodruff_transformation)                                               for reducing the                                               dimensionality of a vector                                               space to an embedded                                               space.<br></div><div> <br></div></div></div><div class=""><br></div><div class="">I think that it                                             would be interesting to add                                             to scipy other methods for                                             subspace embedding (like the                                             Johnson-Lindenstrauss) and                                             build some algorithms on top                                             of it for things like least                                             squeres or low rank                                             approximation.<br></div><div class=""><br></div><div class="">Would some other                                             people be interesting in                                             this?<br></div><div class=""><br></div><div class=""><br></div><div class="">PS: I have a                                             project called <a class="" href="https://github.com/jomsdev/randNLA">RandNLA</a> where I implemented some of                                             the methods of RNLA. The                                             idea is to implement only                                             the most important methods                                             of RNLA in scipy and have                                             this other library for                                             experimenting with new                                             methods and APIs. That will                                             let us not overloading scipy                                             with features if people are                                             not interested in them and                                             focus on the ones that                                             really brings value to the                                             community.<br></div><div class=""><br></div><div class="">Thanks,<br></div><div class=""><br></div><div class="">Jordi.<br></div></div><div class=""><div><br></div><div class=""><div>On 10 January                                             2018 at 10:42, Ralf Gommers <span class=""><<a class="" href="mailto:ralf.gommers@gmail.com">ralf.gommers@gmail.com</a>></span> wrote:<br></div><div> <br></div><blockquote class="" style="margin: 0 0 0 0.8ex; border-left: 1px rgb(204, 204, 204) solid; padding-left: 1ex"><div class="" dir="ltr"><div class=""><div class=""><div class=""><div class=""><div class=""><div>Hi                                                           all,<br></div><div> <br></div></div><div>The GSoC                                                         schedule is a                                                         bit earlier than                                                         normal this                                                         year. The PSF is                                                         asking for ideas                                                         pages to be up                                                         and in decent                                                         shape by Jan                                                         19th. So we'll                                                         need to come up                                                         with some                                                         content quick if                                                         we want to                                                         participate.<br></div><div> <br></div><div> <br></div></div><div>Who is interested                                                       in mentoring this                                                       year?<br></div><div> <br></div><div> <br></div></div><div>I'm happy to do the                                                     admin again, but                                                     probably won't have                                                     time to mentor.<br></div><div> <br></div><div> <br></div></div><div>Cheers,<br></div><div> <br></div></div><div>Ralf<br></div><div> <br></div><div> <br></div></div><div><br></div><div>_______________________________________________<br></div><div> SciPy-Dev mailing list<br></div><div> <a class="" href="mailto:SciPy-Dev@python.org">SciPy-Dev@python.org</a><br></div><div> <a class="" href="https://mail.python.org/mailman/listinfo/scipy-dev">https://mail.python.org/mailman/listinfo/scipy-dev</a><br></div><div> <br></div></blockquote></div><div><br></div></div><div>_______________________________________________<br></div><div> SciPy-Dev mailing list <br></div><div> <a class="" href="mailto:SciPy-Dev@python.org">SciPy-Dev@python.org</a> <br></div><div> <a class="" href="https://mail.python.org/mailman/listinfo/scipy-dev">https://mail.python.org/mailman/listinfo/scipy-dev</a> <br></div></div></blockquote></span></div></div></div></div></div><div><br></div></div><div><br></div></div><div><br></div><div>_______________________________________________<br></div><div> SciPy-Dev mailing list<br></div><div> <a class="" href="mailto:SciPy-Dev@python.org">SciPy-Dev@python.org</a><br></div><div> <a class="" href="https://mail.python.org/mailman/listinfo/scipy-dev">https://mail.python.org/mailman/listinfo/scipy-dev</a><br></div><div> <br></div></blockquote></div><div><br></div></div><div>_______________________________________________<br></div><div> SciPy-Dev mailing list<br></div><div> <a class="" href="mailto:SciPy-Dev@python.org">SciPy-Dev@python.org</a><br></div><div> <a href="https://mail.python.org/mailman/listinfo/scipy-dev" class="moz-txt-link-freetext">https://mail.python.org/mailman/listinfo/scipy-dev</a><br></div></div></blockquote></div><div><br></div></div><div><br></div><div><br></div><pre>_______________________________________________
SciPy-Dev mailing list
<a href="mailto:SciPy-Dev@python.org" class="moz-txt-link-abbreviated">SciPy-Dev@python.org</a>
<a href="https://mail.python.org/mailman/listinfo/scipy-dev" class="moz-txt-link-freetext">https://mail.python.org/mailman/listinfo/scipy-dev</a>
<br></pre></blockquote><div><br></div><div>_______________________________________________<br></div><div>SciPy-Dev mailing list <br></div><div><a href="mailto:SciPy-Dev@python.org">SciPy-Dev@python.org</a> <br></div><div><a href="https://mail.python.org/mailman/listinfo/scipy-dev">https://mail.python.org/mailman/listinfo/scipy-dev</a> <br></div></div></blockquote></div><div><br></div></div><br></body></html>