C++ Logo

SG19

Advanced search

Subject: Re: Jan 9 SG19 Zoom
From: Phil Ratzloff (Phil.Ratzloff_at_[hidden])
Date: 2020-01-09 14:44:34


I'll volunteer to do the submission, if you're OK with that Richard. I know Michael's pretty busy.

-----Original Message-----
From: SG19 <sg19-bounces_at_[hidden]> On Behalf Of Richard Dosselmann via SG19
Sent: Thursday, January 09, 2020 3:38 PM
To: Michael Wong via SG19 <sg19_at_[hidden]>
Cc: dosselmr_at_[hidden]
Subject: Re: [SG19] Jan 9 SG19 Zoom

EXTERNAL

Unfortunately, I was busy with teaching during the meeting today. Are you able to please submit my next version of the stats proposal to the committee for Prague? It is attached to this email.

Thanks,
Richard

Quoting Michael Wong via SG19 <sg19_at_[hidden]>:

> Meeting notes by Michael W
>
> On Thu, Jan 9, 2020 at 12:03 PM Michael Wong <fraggamuffin_at_[hidden]> wrote:
>
>> SG19 Machine Learning 2 hours
>>
>> Hi,
>>
>> Michael Wong is inviting you to a scheduled Zoom meeting.
>>
>> Topic: SG19 monthly Dec 2019-Mar 2020
>> Time: Jan 9 , 2019 01:00 PM Eastern Time (US and Canada)
>> Every month on the Second Thu, until Mar 12, 2020, 4 occurrence(s)
>> Dec 12, 2019 01:00 PM
>> Jan 9, 2020 01:00 PM
>> Feb 13, 2020 01:00 PM
>> Mar 12, 2020 01:00 PM
>> Please download and import the following iCalendar (.ics) files
>> to your calendar system.
>> Monthly:
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fiso
>> .zoom.us%2Fmeeting%2Fu5IudOyrrD0o6QfidlHRJTh2cH9H6gRuPA%2Fics%3FicsTo
>> ken%3D98tyKu-urTgvGdaTslyCe7MtOZX_bN-xjSd_rKpZkDfXKRRbMADeb8oUNYBqIPm
>> B&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa56064994764a08
>> d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C6371419933165
>> 51577&amp;sdata=QdlGs5GNSPYAyJ%2F0akPgccsb3IfpZi7sgv6lovPMfTg%3D&amp;
>> reserved=0
>>
>> Join from PC, Mac, Linux, iOS or Android:
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fiso
>> .zoom.us%2Fj%2F663353262&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7
>> C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%
>> 7C0%7C0%7C637141993316561563&amp;sdata=cWhL5anzul4o0w363BFswBlI8KvQ9Y
>> 6%2BaxfMAto8GuU%3D&amp;reserved=0
>>
>> Or iPhone one-tap :
>> US: +14086380968,,663353262# or +16468769923,,663353262# Or
>> Telephone:
>> Dial(for higher quality, dial a number based on your current
>> location):
>> US: +1 408 638 0968 or +1 646 876 9923 or +1 669 900 6833 or
>> 877
>> 369 0926 (Toll Free) or 877 853 5247 (Toll Free)
>> Meeting ID: 663 353 262
>> International numbers available:
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fiso
>> .zoom.us%2Fu%2FabhaIjFKLZ&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%
>> 7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f
>> %7C0%7C0%7C637141993316561563&amp;sdata=wqEL4Eklg%2BvWI7df%2BbMr%2B2i
>> nlsNEXY5WLdngI63MH%2B4%3D&amp;reserved=0
>>
>> Or Skype for Business (Lync):
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fiso
>> .zoom.us%2Fskype%2F663353262&amp;data=02%7C01%7CPhil.Ratzloff%40sas.c
>> om%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0
>> c2f%7C0%7C0%7C637141993316561563&amp;sdata=K4AdBXSbeBi3XkUc3ZgZuM0PsO
>> EdlOElHTm7awmeJis%3D&amp;reserved=0
>>
>> Agenda:
>>
>> 1. Opening and introductions
>>
>> 1.1 Roll call of participants
>>
> Phil Ratsloff, Marco Foco, Andrew Lumsdaine, jesun Firoz, Kdeweese,
> Michael Wong, Matthew Galati, Vaibhav, william tambellini, Jens Maurer
>
>> 1.2 Adopt agenda
>>
>> 1.3 Approve minutes from previous meeting, and approve publishing
>> previously approved minutes to ISOCPP.org
>>
>> 1.4 Action items from previous meetings
>>
>> 2. Main issues (125 min)
>>
>> 2.1 General logistics
>> All C++ reflector are now moved to listserv
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Flis
>> ts.isocpp.org%2Fmailman%2Flistinfo.cgi%2Fsg19&amp;data=02%7C01%7CPhil
>> .Ratzloff%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c3625
>> 45b3a4309552373a0c2f%7C0%7C0%7C637141993316561563&amp;sdata=lmsmSWJjK
>> tkMyx%2FTPhvV%2F1bLa%2F%2FRiho%2FncApa5Rf8m4%3D&amp;reserved=0
>>
> Lists of features we are working on:
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdocs
> .google.com%2Fspreadsheets%2Fd%2F1JnUJBO72QVURttkKr7gn0_WjP--P0vAne8JB
> fzbRiy0%2Fedit%23gid%3D0&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C
> 66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C
> 0%7C0%7C637141993316561563&amp;sdata=XrPqA3lMOBLJzAlrT8KiuqKIhz0rJOFvx
> KbarrCQ1DA%3D&amp;reserved=0
>
>>
>> 2.2 Paper reviews
>>
>> Review Belfast results.
>>
>> 2.2.1: ML topics
>>
>> Richard Dosselman
>>
>> P1708R1: Math proposal for Machine Learning
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc
>> s.google.com%2Fdocument%2Fd%2F1VAgcyvL1riMdGz7tQIT9eTtSSfV3CoCEMWKk8G
>> vVuFY%2Fedit&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa560
>> 64994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C63
>> 7141993316561563&amp;sdata=wgwXlGryOe90tTvhx3DG6fK56NuKIb5JYLqR%2FVYW
>> 0XU%3D&amp;reserved=0
>>
>> P1709R1: Graph Proposal for Machine Learning
>>
> R2 proposal
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdocs
> .google.com%2Fspreadsheets%2Fd%2F1JnUJBO72QVURttkKr7gn0_WjP--P0vAne8JB
> fzbRiy0%2Fedit%23gid%3D0&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C
> 66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C
> 0%7C0%7C637141993316561563&amp;sdata=XrPqA3lMOBLJzAlrT8KiuqKIhz0rJOFvx
> KbarrCQ1DA%3D&amp;reserved=0
> have 2 graph data structures
> 1 for directed and 1 for ordered graph last telecon we show DFS and
> BFS implemented as ranges
>
> what we have done
> also have undirected graphs - what is the difference between directed
> and undirected?
> - both have vertex and edges, but undirected edges are edges, for
> directed there are out and in edges, was trying to use a uniform api
> to remove that distinction
> - still one place we use in and out where edge refers to 2 vertices
> implemented shortest paths, also have transitive closure I have added
> category types, on whether we have sparse or dense graph, a directed
> graph is a uniform graph and an out directed graph bidirected, and
> unidirected follows a sparse graph and a vertex range request
> definition of dense (a matrix, random access inner, forward iterator
> outter) vs sparse graph (random access outter, forward iterator to
> inner)
> we already have existing algo in library for textsearches propose
> remove sparse and dense graph as it is implementation details why
> integral vertex key? internal implementation requires we have integer
> key to look up vertex, vector in this case, want to relax overtime the
> less restrictive one is random access iterator, so a DAG has random
> access iterators what is teh key? -a DAG the key is from the vertex of
> container, for a map the key is defined by user should contrain on
> OutIter result_iter please need to specify restrictions on G and so
> use concepts, same with OutIter
>
> shortest path has arbitrary list so has to do memory allocation
> shortest distance has no memory allocation
> BGL17 approach didnt really have shortest path or BFS do anything
> other then go through the graph when u get to last vertex for the
> path, want to iterate each vertexes of the path, so have inner and
> outer loop you need to walk through
>
> will you output distance to targets with a filter? No there is no
> filter shortest distance and shortest paths give you the same opotion
> with leaves_only, so if you want mre sophisticated filters, it is
> possible Do you really need leaves_only as a parameter; shortest_paths
> would only need the parent, any partial path is also the shortest
> path, for Dijkstra, just store the parent pointer instead of
> constructing the paths so give them a range of what we have now, with
> the distance may be a better way of doing that another output u may
> want is an edge iterator, using a pair , along with a vertex iterator
>
> parent and distance, may want a path of vertexes directly; instead of
> returning a vector of things, we just return a range ok I will go back
> and work on getting both, implementation later after the mailing
> deadline
>
> what examples of graph data structure should we have?
> compressed sparse row (CSR) is useful, for the user to build their own
> concern that we are interpreting graph as ranges of ranges,notion of
> dereferecing can be cumbersome, - that depends on on what algorithm
> actually needs, so a compressed data structure is useful to meet the
> need of the algorithm, so you can use MKL under the hood, may need to
> sync with Intel, graph-BLAS as an extension of MKL, using CSR
> optimizing for AVX sparse LA or graphs are memory bound, theer is not
> any optimization you can apply to move things in memory any faster CSR
> and adjacency edge list discussion approach Intel Moscow team to see
> how they want to participate
>
> in our case, we are not distinguishign dense vs sparse graph based on
> Jens comment
>
> adapting this to their own data structues, - yes Jens mention wantign
> interface to be minimal ,if std library has equivalent things, then we
> shoudl reuse them.
> I will look at CSR options
>
>
> R1 proposal
>
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc
>> s.google.com%2Fdocument%2Fd%2F1QkfDzGyfNQKs86y053M0YHOLP6frzhTJqzg1Ug
>> _vkkE%2Fedit%3Fusp%3Dsharing&amp;data=02%7C01%7CPhil.Ratzloff%40sas.c
>> om%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0
>> c2f%7C0%7C0%7C637141993316561563&amp;sdata=5ePNdrozcJeDyPEbdWTEJuEzBz
>> CujHmo5avWSZp1NmY%3D&amp;reserved=0
>> <
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc
>> s.google.com%2Fdocument%2Fd%2F1QkfDzGyfNQKs86y053M0YHOLP6frzhTJqzg1Ug
>> _vkkE%2Fedit%3Fusp%3Dsharing&amp;data=02%7C01%7CPhil.Ratzloff%40sas.c
>> om%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0
>> c2f%7C0%7C0%7C637141993316561563&amp;sdata=5ePNdrozcJeDyPEbdWTEJuEzBz
>> CujHmo5avWSZp1NmY%3D&amp;reserved=0>
>>
>>
>> P1707R0: Differentiable Programming by Marco Foco
>>
> explain differences automatic, vs symbolic differentiation
> - auto works on algorithm ,and uses teh structure of the algorithm
> - sym only works on expressions
> also motivate why we want langage solution and not a library solution
> -performance example demonstrate benefits -but mainly you cannot do
> many things with library, need to use a specially named type as in
> Enoki want to enable this auto diff tool kit for other things like
> matrices matrix and vector types support is very important assume
> people using this library is different then the library writer - so we
> will often combine multiple libraries together
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Feige
> n.tuxfamily.org%2Fdox%2Funsupported%2Fgroup__AutoDiff__Module.html&amp
> ;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa56064994764a08d79543
> d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C637141993316561563&a
> mp;sdata=MYjAEZ4TQ5rbS7hF3pTp6SgiQY%2BBIkSw9GLmdPg9yj4%3D&amp;reserved
> =0
> prompt Andrew L on Tulsa library
> example:
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fjoel
> cfd.com%2Fautomatic-differentiation%2F&amp;data=02%7C01%7CPhil.Ratzlof
> f%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309
> 552373a0c2f%7C0%7C0%7C637141993316561563&amp;sdata=g7T3IV4XE4C3hLJoB7U
> l2YTjpL7l9ZaNYM3CyfK0F5M%3D&amp;reserved=0
>
> a fork of Eigen "aimed at Algorithmic Differentiation (AD) ":
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgitl
> ab.stce.rwth-aachen.de%2Fstce%2Feigen-ad&amp;data=02%7C01%7CPhil.Ratzl
> off%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a43
> 09552373a0c2f%7C0%7C0%7C637141993316561563&amp;sdata=5JLtBI7gmm0NjTA9o
> yKKUdU%2BqnHPzhSxEq5McbXlUcM%3D&amp;reserved=0
> A recent paper:
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Farxi
> v.org%2Fabs%2F1911.12604&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C
> 66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C
> 0%7C0%7C637141993316561563&amp;sdata=FaYgbhMVPxTuWw3B6uMNbggSlvq17Ydt3
> jmzUSbIH94%3D&amp;reserved=0
>
>
>
> Numerical Differentiation - a new paper Could be a library solution
> which can land earlier limit and choosing h sqrt in C++20 is still not
> constexpr is there a bets practice for moving instead of copying a
> function when it is a pure function: let lewg decide e.g. matrix
> solver ,so good to memoize f(x), and repeatedly call it with different
> h AL to send reference complex step is not really a numerical solution
>
>> P1416R1: SG19 - Linear Algebra for Data Science and Machine Learning
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc
>> s.google.com%2Fdocument%2Fd%2F1IKUNiUhBgRURW-UkspK7fAAyIhfXuMxjk7xKik
>> K4Yp8%2Fedit%23heading%3Dh.tj9hitg7dbtr&amp;data=02%7C01%7CPhil.Ratzl
>> off%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4
>> 309552373a0c2f%7C0%7C0%7C637141993316561563&amp;sdata=OY%2BYs3zVV4oe0
>> 42udx9%2Fpy8YYDxoGwLjbVwg5Z0jTso%3D&amp;reserved=0
>>
>> P1415: Machine Learning Layered list
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc
>> s.google.com%2Fdocument%2Fd%2F1elNFdIXWoetbxjO1OKol_Wj8fyi4Z4hogfj5tL
>> VSj64%2Fedit%23heading%3Dh.tj9hitg7dbtr&amp;data=02%7C01%7CPhil.Ratzl
>> off%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4
>> 309552373a0c2f%7C0%7C0%7C637141993316571559&amp;sdata=JacY0mDYX9g7nDi
>> rVJKAgIb%2FNMX6ZKN%2B8nrFxgsywXc%3D&amp;reserved=0
>>
>> 2.2.2 SG14 Linear Algebra progress:
>> Different layers of proposal
>>
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fdoc
>> s.google.com%2Fdocument%2Fd%2F1poXfr7mUPovJC9ZQ5SDVM_1Nb6oYAXlK_d0ljd
>> UAtSQ%2Fedit&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa560
>> 64994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C63
>> 7141993316571559&amp;sdata=mVP%2FPVYdZ4AOeVuG6gQBwYbxrF%2Fh%2BQ5watNx
>> tj1XlG8%3D&amp;reserved=0
>>
>> 2.2.3 any other proposal for reviews?
>>
>> 2.3 Other Papers and proposals
>>
>> 2.5 Future F2F meetings:
>>
> Paper submission:
> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fisoc
> pp.org%2Fpapers&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa5
> 6064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C6
> 37141993316571559&amp;sdata=tUng86opftHV2VPkOKICsmU6NlnGC7BCIdfRMT6Seb
> 0%3D&amp;reserved=0
>
>> 2.6 future C++ Standard meetings:
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fiso
>> cpp.org%2Fstd%2Fmeetings-and-participation%2Fupcoming-meetings&amp;da
>> ta=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa56064994764a08d79543d4
>> 29%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C637141993316571559&am
>> p;sdata=ijTPGyG4mnZXtyGN57LuTfAjhDhS97WrZ3Cg6x%2BFMIU%3D&amp;reserved
>> =0
>>
>> -2020-02-10 to 15: Prague, Czech Republic
>>
>> - 2020-06-01 to 06: Bulgaria
>> - 2020-11: (New York, tentative)
>> - 2021-02-22 to 27: Kona, HI, USA
>>
>> 3. Any other business
>>
>> New reflector
>>
>> https://nam02.safelinks.protection.outlook.com/?url=http%3A%2F%2Flist
>> s.isocpp.org%2Fmailman%2Flistinfo.cgi%2Fsg19&amp;data=02%7C01%7CPhil.
>> Ratzloff%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c36254
>> 5b3a4309552373a0c2f%7C0%7C0%7C637141993316571559&amp;sdata=6sg82vD57k
>> QsSGy%2BiHkfaZP%2FxCL2blBg0ns2D4VAwyw%3D&amp;reserved=0
>>
>> Old Reflector
>> https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgro
>> ups.google.com%2Fa%2Fisocpp.org%2Fforum%2F%23!newtopic%2Fsg19&amp;dat
>> a=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa56064994764a08d79543d42
>> 9%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C637141993316571559&amp
>> ;sdata=aI4X3ReTZSkuIcga%2BHJhX%2FkAzjpVBPba2dm1i4pwoNE%3D&amp;reserve
>> d=0
>> <https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgr
>> oups.google.com%2Fa%2Fisocpp.org%2Fforum%2F%3Ffromgroups%3D%23!forum%
>> 2Fsg14&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa560649947
>> 64a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C63714199
>> 3316571559&amp;sdata=wUuqqc9x4P3qrYKSwjS10y1V7elgNAJVbTfMmlAoVr0%3D&a
>> mp;reserved=0>
>>
>> Code and proposal Staging area
>>
>> 4. Review
>>
>> 4.1 Review and approve resolutions and issues [e.g., changes to SG's
>> working draft]
>>
>> 4.2 Review action items (5 min)
>>
>> 5. Closing process
>>
>> 5.1 Establish next agenda
>>
>> TBD
>>
>> 5.2 Future meeting
>>
>> Dec 12, 2019 01:00 PM
>> Jan 9, 2020 01:00 PM: Jan 12 is mailing deadline
>> Feb 13, 2020 01:00 PM
>> Mar 12, 2020 01:00 PM
>>
>

-----------------------------------------------------
This e-mail was sent via the Secure Web Server at the University of Regina, Department of Computer Science
          https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.cs.uregina.ca%2FWebMail%2F&amp;data=02%7C01%7CPhil.Ratzloff%40sas.com%7C66a5c1fa56064994764a08d79543d429%7Cb1c14d5c362545b3a4309552373a0c2f%7C0%7C0%7C637141993316571559&amp;sdata=KcONaeE8xsQ7IROcOITPhpcIlerhn2m3RKNY9IheQ3I%3D&amp;reserved=0


SG19 list run by sg19-owner@lists.isocpp.org

SG19 Archives on Google Groups