13:54:27 RRSAgent has joined #webmachinelearning 13:54:27 logging to https://www.w3.org/2019/05/09-webmachinelearning-irc 13:54:31 Zakim has joined #webmachinelearning 13:54:36 RRSAgent, make logs public 13:54:42 Meeting: WebML CG Teleconference – 9 May 2019 13:54:48 Chair: Anssi 13:55:03 Scribe: Anssi 13:55:09 Agenda: https://github.com/webmachinelearning/meetings/blob/master/telcons/2019-05-09-agenda.md 13:55:20 scribeNick: anssik 13:55:27 Present+ Anssi_Kostiainen 13:55:33 RRSAgent, draft minutes v2 13:55:33 I have made the request to generate https://www.w3.org/2019/05/09-webmachinelearning-minutes.html anssik 13:56:58 tomoyuki has joined #webmachinelearning 13:58:48 Rama has joined #webmachinelearning 13:59:28 belem has joined #webmachinelearning 13:59:30 ningxin_hu has joined #webmachinelearning 14:00:32 Present+ Ningxin_Hu 14:00:43 Present+ Tomoyuki_Shimizu 14:00:49 Present+ Rafael_Cintron 14:00:55 Present+ Ganesan_Ramalingam 14:01:01 Rafael has joined #webmachinelearning 14:01:04 kainino has joined #webmachinelearning 14:01:15 Present+ Kai_Ninomiya 14:01:36 RRSAgent, draft minutes v2 14:01:36 I have made the request to generate https://www.w3.org/2019/05/09-webmachinelearning-minutes.html anssik 14:01:48 TOPIC: Graph-building syntax simpler for web developers 14:02:05 -> https://github.com/webmachinelearning/webnn/issues/16 Graph-building syntax simpler for web developers #16 14:02:07 Present+ Greg Whitworth 14:05:34 rama: proposal to hide implementation details, i.e. that the operands are represented by integers 14:06:46 ningxin_hu: agree proposal (1) from rama is a good improvement for ergonomics 14:14:45 rama: addOperation can create and return an result operators it creates 14:15:29 ningxin_hu: for (2) feedback is, how to support optional output of tensors, batch normalization etc. 14:18:32 ningxin_hu: addition comment, as mentioned, foundation spec is direct derivative from Android NN API which is a C style API, inherits its style 14:19:03 ... due to this background, my suggestion is to survey other graph-building APIs in native ecosystem to learn API design patterns 14:21:33 Present+ Paul_McDaniel 14:22:20 paul: would love to contribute our learnings from two Microsoft's graph-building APIs 14:24:18 PROPOSED RESOLUTION: open a separate issue for surveying other graph-building APIs from native ecosystem 14:24:28 RESOLUTION: open a separate issue for surveying other graph-building APIs from native ecosystem 14:24:42 TOPIC: Define the set of operations and their specification 14:24:52 -> https://github.com/webmachinelearning/webnn/issues/17 Define the set of operations and their specification 14:26:00 gregwhitworth: don't want to duplicate effort, many folks representing ONNX on the call who have done definition of operations 14:26:42 ... prefer keep us focused on JS API and reference pre-existing other spec for a set of operations 14:26:44 RRSAgent, draft minutes v2 14:26:44 I have made the request to generate https://www.w3.org/2019/05/09-webmachinelearning-minutes.html anssik 14:29:20 kainino: not working on TF, not able to give Google's position 14:29:58 Present+ Jonathan_Bingham 14:31:36 Jonathan: heard a concern internally, that not sure if by depending on external reference we can keep up with rapidly developing set of ops 14:32:31 Jonathan has joined #webmachinelearning 14:34:07 dom has joined #webmachinelearning 14:34:29 Jonathan: I'm volunteering to take an action to find out internally at Google whether referencing ONNX would be reasonable 14:35:24 paul: thanks for digging in 14:36:45 ningxin_hu: as mentioned in this issue, in our POC we've studied TF Lite and ONNX model support 14:37:43 ... we've looked at models for use cases identified by this group, there's a spreadsheet with mapping 14:38:18 ... maybe could start with a subset of ops driven by use cases 14:39:21 https://docs.google.com/spreadsheets/d/1nthZOwgIKsj34EB-SymEwoNTPsxo4X8Pxavm-JaBwME/edit?usp=sharing 14:42:21 anssik: how is the ONNX set of operations evolved, what is the process? 14:42:49 paul: use case driven, surveyed existing frameworks, common denominator 14:42:53 this is the models and required op spreadsheet: https://docs.google.com/spreadsheets/d/1S4WOyyQpAQy7Ac56KVTApv-M691ItTLpFzrrH7aAi6Y/edit?usp=sharing 14:43:00 ... evolving the set is community-driven in the open 14:43:41 ... releases every ~2 months 14:45:21 s/https://docs.google.com/spreadsheets/d/1nthZOwgIKsj34EB-SymEwoNTPsxo4X8Pxavm-JaBwME/edit?usp=sharing/(removed) 14:45:50 TOPIC: W3C Workshop on Web & Machine Learning 14:46:18 -> https://www.w3.org/2018/11/machine-learning-workshop/ W3C Workshop on Web & Machine Learning 14:47:18 dom: W3C Staff, working with Anssi on many other projects e.b. WebVR 14:47:36 ... a year ago started project how to use ML in the context of web browsers 14:48:37 RRSAgent, draft minutes v2 14:48:37 I have made the request to generate https://www.w3.org/2019/05/09-webmachinelearning-minutes.html anssik 14:49:06 ... target of the workshop would be to bring the broader community together 14:54:14 dom: would be useful to get input on dates 14:54:19 ... what events to avoid clash with 14:55:40 paul: not aware of any October dates that would block 14:57:06 rafael: no bandwidth to organize, but +1 to participate if dates work 14:59:10 RRSAgent, draft minutes v2 14:59:10 I have made the request to generate https://www.w3.org/2019/05/09-webmachinelearning-minutes.html anssik 14:59:14 dom@w3.org 15:00:45 TOPIC: Adjust 15:00:48 RRSAgent, draft minutes v2 15:00:48 I have made the request to generate https://www.w3.org/2019/05/09-webmachinelearning-minutes.html anssik 15:00:57 belem has left #webmachinelearning 16:49:42 Zakim has left #webmachinelearning