From 7ef48e78ab8544c542ffacf6fab09553a0ae03b7 Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Thu, 3 Sep 2020 21:57:49 +0300 Subject: [PATCH 1/7] Add workshop agenda --- agenda/README.md | 43 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) create mode 100644 agenda/README.md diff --git a/agenda/README.md b/agenda/README.md new file mode 100644 index 0000000..f3763a7 --- /dev/null +++ b/agenda/README.md @@ -0,0 +1,43 @@ +# W3C Workshop on Web and Machine Learning + +![Agenda header image][header] + +# Agenda for Live Sessions + +## September 16, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200916T14) + +>πŸ”Ž **Scope:** Opportunities and Challenges of Browser-Based Machine Learning +> +>βœ… **Goal:** Determine what are the unique opportunities of browser-based ML, what are the obstacles hindering adoption + +- πŸ‘‹ Introductions +- [πŸ’‘ "Opportunities and Challenges" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Opportunities+and+Challenges%22+sort%3Acomments-desc) + + +## September 22, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200922T14) + +>πŸ”Ž **Scope:** Web Platform Foundations for Machine Learning +> +>βœ… **Goal:** Goal: Understand how machine learning fits into the Web technology stack + +- [πŸ’‘ "Web Platform Foundations" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Web+Platform+Foundations%22+sort%3Acomments-desc) + +## September 23, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200923T14) + +>πŸ”Ž **Scope:** Machine Learning Experiences on the Web: A **Developer’s** Perspective +> +>βœ… **Goal:** Authoring ML experiences on the Web; challenges and opportunities of reusing existing ML models on the Web; on-device training, known technical solutions, gaps + +- [πŸ’‘ "Developer's Perspective" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Developer%27s+Perspective%22+sort%3Acomments-desc) + +## September 29, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200929T14) + +>πŸ”Ž **Scope:** Machine Learning Experiences on the Web: A **User’s** Perspective +> +>βœ… **Goal:** Goal: Web & ML for all: education, learning, accessibility, cross-industry experiences, cross-disciplinary ML: music, art, and media meet ML; Share learnings and best practices across industries + +- [πŸ’‘ "User's Perspective" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22User%27s+Perspective%22+sort%3Acomments-desc) +- πŸ‘ Conclusions & Next Steps + + +[header]: https://www.w3.org/comm/assets/graphics/web-ML/workshop-bg.png "W3C Workshop on Web and Machine Learning header" From 74381d70231a565288e5d622e810dc86fdf2046d Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Fri, 4 Sep 2020 11:37:25 +0300 Subject: [PATCH 2/7] Adjust introduction scope --- agenda/README.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/agenda/README.md b/agenda/README.md index f3763a7..302231f 100644 --- a/agenda/README.md +++ b/agenda/README.md @@ -10,7 +10,7 @@ > >βœ… **Goal:** Determine what are the unique opportunities of browser-based ML, what are the obstacles hindering adoption -- πŸ‘‹ Introductions +- πŸ‘‹ Introduction to the workshop - [πŸ’‘ "Opportunities and Challenges" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Opportunities+and+Challenges%22+sort%3Acomments-desc) @@ -18,7 +18,7 @@ >πŸ”Ž **Scope:** Web Platform Foundations for Machine Learning > ->βœ… **Goal:** Goal: Understand how machine learning fits into the Web technology stack +>βœ… **Goal:** Understand how machine learning fits into the Web technology stack - [πŸ’‘ "Web Platform Foundations" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Web+Platform+Foundations%22+sort%3Acomments-desc) @@ -34,7 +34,7 @@ >πŸ”Ž **Scope:** Machine Learning Experiences on the Web: A **User’s** Perspective > ->βœ… **Goal:** Goal: Web & ML for all: education, learning, accessibility, cross-industry experiences, cross-disciplinary ML: music, art, and media meet ML; Share learnings and best practices across industries +>βœ… **Goal:** Web & ML for all: education, learning, accessibility, cross-industry experiences, cross-disciplinary ML: music, art, and media meet ML; Share learnings and best practices across industries - [πŸ’‘ "User's Perspective" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22User%27s+Perspective%22+sort%3Acomments-desc) - πŸ‘ Conclusions & Next Steps From dc494a9feab22b055de26908fc018a1cc9b28a55 Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Mon, 7 Sep 2020 22:11:30 +0300 Subject: [PATCH 3/7] Add "Opportunities and Challenges" discussion outline --- agenda/README.md | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) diff --git a/agenda/README.md b/agenda/README.md index 302231f..2e4aeb4 100644 --- a/agenda/README.md +++ b/agenda/README.md @@ -10,9 +10,33 @@ > >βœ… **Goal:** Determine what are the unique opportunities of browser-based ML, what are the obstacles hindering adoption -- πŸ‘‹ Introduction to the workshop -- [πŸ’‘ "Opportunities and Challenges" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Opportunities+and+Challenges%22+sort%3Acomments-desc) +πŸ‘‹ Introduction to the workshop - @anssiko @dontcallmedom +[πŸ’‘ "Opportunities and Challenges" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Opportunities+and+Challenges%22+sort%3Acomments-desc): + +ℹ️ [WebGPU fitness for ML frameworks #66](https://github.com/w3c/machine-learning-workshop/issues/66) - @jasonmayes @Kangz @grorg +- ❓ Does WebGPU expose the right API surface to support ML frameworks interactions with GPUs? +- βœ”οΈ Proposal: New WebGPU extensions for subgroups, cooperative matrix multiply. + +ℹ️ [Applicability to non-browser JS environments #62](https://github.com/w3c/machine-learning-workshop/issues/62) - @jasonmayes @phoddie @huningxin @WenheLI +- ❓ Pay attention to the applicability of the browser-targeted work to non-browser JS environments, in particular Node.js. +- βœ”οΈ Proposal: Extend W3C coordination to TC53 and non-browser projects. + +ℹ️ [Protecting ML models #67](https://github.com/w3c/machine-learning-workshop/issues/67) - @jasonmayes @tidoust @pyu10055 @jbingham +- ❓ Some ML providers need to ensure their ML models cannot be extracted from a browser app. +- βœ”οΈ Proposal: Investigate existing access control mechanisms for video, learnings from 3D assets. + +ℹ️ [Support for Float16 in JS & Wasm environments #64](https://github.com/w3c/machine-learning-workshop/issues/64) - @cynthia @jasonmayes +- ❓ Lack of support for float16 in JS and Wasm environments problematic for quantized models. +- βœ”οΈ Proposal: TBD + +ℹ️ [In-browser training #82](https://github.com/w3c/machine-learning-workshop/issues/82) and [Training across devices #83](https://github.com/w3c/machine-learning-workshop/issues/83) - @irealva @cynthia +- ❓ The current in-browser efforts are focused on inference rather than training. +- βœ”οΈ Proposal: Understand successful real-world usages (e.g. Teachable Machine) and target transfer learning as the initial training use case for related browser API work. + +ℹ️ [Permission model for Machine Learning APIs #72](https://github.com/w3c/machine-learning-workshop/issues/72) - @cynthia @dontcallmedom @anssiko +- ❓ How to design a forward-looking permission model for ML APIs? +- βœ”οΈ Proposal: TBD ## September 22, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200922T14) From 3912ae9406d395c864e9d7864b454cdb841ef044 Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Tue, 8 Sep 2020 17:08:25 +0300 Subject: [PATCH 4/7] Add memory copies to "Opportunities and Challenges" --- agenda/README.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/agenda/README.md b/agenda/README.md index 2e4aeb4..6bdb458 100644 --- a/agenda/README.md +++ b/agenda/README.md @@ -34,6 +34,10 @@ - ❓ The current in-browser efforts are focused on inference rather than training. - βœ”οΈ Proposal: Understand successful real-world usages (e.g. Teachable Machine) and target transfer learning as the initial training use case for related browser API work. +ℹ️ [Memory copies #93](https://github.com/w3c/machine-learning-workshop/issues/93) - @aboba @wchao1115 +- ❓ Machine learning apps within the browser using the media pipeline trigger many more memory copies compared with native applications hindering performance. +- βœ”οΈ Proposal: Introduce a more direct way to feed a video frame, possibly captured from a camera, to a ML model. + ℹ️ [Permission model for Machine Learning APIs #72](https://github.com/w3c/machine-learning-workshop/issues/72) - @cynthia @dontcallmedom @anssiko - ❓ How to design a forward-looking permission model for ML APIs? - βœ”οΈ Proposal: TBD From 10b61332165edab39413df97ec132c1271530b2d Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Thu, 10 Sep 2020 10:25:21 +0300 Subject: [PATCH 5/7] Add topical organization for Sep 16 --- agenda/README.md | 36 +++++++++++++++++++++++++----------- 1 file changed, 25 insertions(+), 11 deletions(-) diff --git a/agenda/README.md b/agenda/README.md index 6bdb458..8957c18 100644 --- a/agenda/README.md +++ b/agenda/README.md @@ -14,34 +14,48 @@ [πŸ’‘ "Opportunities and Challenges" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Opportunities+and+Challenges%22+sort%3Acomments-desc): +## Improving existing web platform capabilities + ℹ️ [WebGPU fitness for ML frameworks #66](https://github.com/w3c/machine-learning-workshop/issues/66) - @jasonmayes @Kangz @grorg - ❓ Does WebGPU expose the right API surface to support ML frameworks interactions with GPUs? - βœ”οΈ Proposal: New WebGPU extensions for subgroups, cooperative matrix multiply. +ℹ️ [Support for Float16 in JS & Wasm environments #64](https://github.com/w3c/machine-learning-workshop/issues/64) - @cynthia @jasonmayes +- ❓ Lack of support for float16 in JS and Wasm environments problematic for quantized models. +- βœ”οΈ Proposal: TBD + +ℹ️ [Memory copies #93](https://github.com/w3c/machine-learning-workshop/issues/93) - @aboba @wchao1115 +- ❓ Machine learning apps within the browser using the media pipeline trigger many more memory copies compared with native applications hindering performance. +- βœ”οΈ Proposal: Introduce a more direct way to feed a video frame, possibly captured from a camera, to a ML model. + +ℹ️ [Permission model for Machine Learning APIs #72](https://github.com/w3c/machine-learning-workshop/issues/72) - @cynthia @dontcallmedom @anssiko +- ❓ How to design a forward-looking permission model for ML APIs? +- βœ”οΈ Proposal: TBD + +## Extending beyond the browser + ℹ️ [Applicability to non-browser JS environments #62](https://github.com/w3c/machine-learning-workshop/issues/62) - @jasonmayes @phoddie @huningxin @WenheLI - ❓ Pay attention to the applicability of the browser-targeted work to non-browser JS environments, in particular Node.js. - βœ”οΈ Proposal: Extend W3C coordination to TC53 and non-browser projects. +ℹ️ [Targeting WASI-NN and WebNN together #96](https://github.com/w3c/machine-learning-workshop/issues/96) - @mehmetoguzderin @mingqiusun @abrown +- ❓ Should libraries for browsers and/or Wasm execution environments be able to target WebNN and WASI-NN together? +- βœ”οΈ Proposal: TBD + +## Considerations for creating and deploying models + ℹ️ [Protecting ML models #67](https://github.com/w3c/machine-learning-workshop/issues/67) - @jasonmayes @tidoust @pyu10055 @jbingham - ❓ Some ML providers need to ensure their ML models cannot be extracted from a browser app. - βœ”οΈ Proposal: Investigate existing access control mechanisms for video, learnings from 3D assets. -ℹ️ [Support for Float16 in JS & Wasm environments #64](https://github.com/w3c/machine-learning-workshop/issues/64) - @cynthia @jasonmayes -- ❓ Lack of support for float16 in JS and Wasm environments problematic for quantized models. -- βœ”οΈ Proposal: TBD +ℹ️ [ML model format #74](https://github.com/w3c/machine-learning-workshop/issues/74) - @cynthia @jbingham @wchao1115 +- ❓ There is no standard format for packaging and shipping ML models, model formats evolve rapidly. +- βœ”οΈ Proposal: Initially focus on defining a Web API for accelerating established reusable ML operations instead of standardizing a model format. ℹ️ [In-browser training #82](https://github.com/w3c/machine-learning-workshop/issues/82) and [Training across devices #83](https://github.com/w3c/machine-learning-workshop/issues/83) - @irealva @cynthia - ❓ The current in-browser efforts are focused on inference rather than training. - βœ”οΈ Proposal: Understand successful real-world usages (e.g. Teachable Machine) and target transfer learning as the initial training use case for related browser API work. -ℹ️ [Memory copies #93](https://github.com/w3c/machine-learning-workshop/issues/93) - @aboba @wchao1115 -- ❓ Machine learning apps within the browser using the media pipeline trigger many more memory copies compared with native applications hindering performance. -- βœ”οΈ Proposal: Introduce a more direct way to feed a video frame, possibly captured from a camera, to a ML model. - -ℹ️ [Permission model for Machine Learning APIs #72](https://github.com/w3c/machine-learning-workshop/issues/72) - @cynthia @dontcallmedom @anssiko -- ❓ How to design a forward-looking permission model for ML APIs? -- βœ”οΈ Proposal: TBD - ## September 22, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200922T14) >πŸ”Ž **Scope:** Web Platform Foundations for Machine Learning From 73d33886ae643dec0be0fd0f33e59cdfb5099bad Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Thu, 17 Sep 2020 16:41:06 +0300 Subject: [PATCH 6/7] Add workshop agenda for Live Sessions #2 and #3 --- agenda/README.md | 62 +++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 53 insertions(+), 9 deletions(-) diff --git a/agenda/README.md b/agenda/README.md index 8957c18..5d006f4 100644 --- a/agenda/README.md +++ b/agenda/README.md @@ -38,9 +38,13 @@ - ❓ Pay attention to the applicability of the browser-targeted work to non-browser JS environments, in particular Node.js. - βœ”οΈ Proposal: Extend W3C coordination to TC53 and non-browser projects. -ℹ️ [Targeting WASI-NN and WebNN together #96](https://github.com/w3c/machine-learning-workshop/issues/96) - @mehmetoguzderin @mingqiusun @abrown -- ❓ Should libraries for browsers and/or Wasm execution environments be able to target WebNN and WASI-NN together? -- βœ”οΈ Proposal: TBD +## September 22, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200922T14) + +>πŸ”Ž **Scope:** Web Platform Foundations for Machine Learning +> +>βœ… **Goal:** Understand how machine learning fits into the Web technology stack + +- [πŸ’‘ "Web Platform Foundations" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Web+Platform+Foundations%22+sort%3Acomments-desc) ## Considerations for creating and deploying models @@ -52,17 +56,23 @@ - ❓ There is no standard format for packaging and shipping ML models, model formats evolve rapidly. - βœ”οΈ Proposal: Initially focus on defining a Web API for accelerating established reusable ML operations instead of standardizing a model format. -ℹ️ [In-browser training #82](https://github.com/w3c/machine-learning-workshop/issues/82) and [Training across devices #83](https://github.com/w3c/machine-learning-workshop/issues/83) - @irealva @cynthia +ℹ️ [In-browser training #82](https://github.com/w3c/machine-learning-workshop/issues/82) - @irealva @hapticdata @cynthia - ❓ The current in-browser efforts are focused on inference rather than training. - βœ”οΈ Proposal: Understand successful real-world usages (e.g. Teachable Machine) and target transfer learning as the initial training use case for related browser API work. -## September 22, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200922T14) + [Training across devices #83](https://github.com/w3c/machine-learning-workshop/issues/83) - @wmaass @Nov1102 @EmmaNingMS @zolkis @jaykishigami +- ❓ Understand the role of edge computing in training and interactions with the web platform. +- βœ”οΈ Proposal: Work with Web & Networks IG to understand edge computing use cases and ensure input from ML usages is considered. ->πŸ”Ž **Scope:** Web Platform Foundations for Machine Learning -> ->βœ… **Goal:** Understand how machine learning fits into the Web technology stack +## Extending the web foundations for ML -- [πŸ’‘ "Web Platform Foundations" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Web+Platform+Foundations%22+sort%3Acomments-desc) +ℹ️ [Targeting WASI-NN and WebNN together #96](https://github.com/w3c/machine-learning-workshop/issues/96) - @mehmetoguzderin @mingqiusun @abrown +- ❓ Should libraries for browsers and/or Wasm execution environments be able to target WebNN and WASI-NN together? +- βœ”οΈ Proposal: TBD + +ℹ️ [Heterogeneous parallel computing for the web #92](https://github.com/w3c/machine-learning-workshop/issues/92) - @jeffhammond @Kangz @abrown +- ❓ How do the heterogeneous parallel computing abstractions fit in with the web platform? +- βœ”οΈ Proposal: TBD ## September 23, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200923T14) @@ -72,6 +82,40 @@ - [πŸ’‘ "Developer's Perspective" discussion topics](https://github.com/w3c/machine-learning-workshop/issues?q=is%3Aissue+is%3Aopen+label%3A%22Developer%27s+Perspective%22+sort%3Acomments-desc) +## Applying web design principles to ML + +ℹ️ [Progressive Enhancement / Graceful degradation #68](https://github.com/w3c/machine-learning-workshop/issues/68) - @dontcallmedom @jbingham @wchao1115 @huningxin +- ❓ How to bring more ML features as optional improvements on more powerful devices and browsers without breaking web compatibility? +- βœ”οΈ Proposal: TBD + +ℹ️ [Conformance testing of ML APIs for the Web #80](https://github.com/w3c/machine-learning-workshop/issues/80) - @wchao1115 @Kangz +- ❓ Robust conformance testing is a cornerstone of the interoperable web platform, how to scale that to the ML APIs and formats? +- βœ”οΈ Proposal: TBD + +## Improving web developer ergonomics + +ℹ️ [JS Operator overloading for Machine Learning #73](https://github.com/w3c/machine-learning-workshop/issues/73) - @TBD +- ❓ Limitations in ECMAScript expressiveness impose ergonomics limitations for JS APIs on the web platform e.g. in vector matrix or tensor operations. +- βœ”οΈ Proposal: TBD + +ℹ️ [WebGL garbage collection #63](https://github.com/w3c/machine-learning-workshop/issues/63) - @jasonmayes @Kangz @wchao1115 @huningxin +- ❓ Garbage collection in the WebGL API affects multiple ML libraries. +- βœ”οΈ Proposal: Identify any improvements in graphics APIs to alleviate the GC issue, ensure purpose-built APIs designed around computational graph abstraction (e.g. WebNN) optimize GC from library usage perspective. + +ℹ️ [Neural network-oriented graph database #102](https://github.com/w3c/machine-learning-workshop/issues/102) - @WenheLI +- ❓ Understand model storage issues on the client, research the feasibility of a neural network-oriented graph database for the web. +- βœ”οΈ Proposal: TBD + +## Developing interactive web experiences with ML + +ℹ️ [Action-Response Cycle bottlenecks in interactive music apps #97](https://github.com/w3c/machine-learning-workshop/issues/97) - @teropa +- ❓ Action-Response Cycle in interactive (music) apps must execute within 20 ms. +- βœ”οΈ Proposal: Investigate inference in AudioWorklet context and media integration e.g. fast streaming inputs from MediaStream. + +ℹ️ Noise suppression with DSP+DNN, WebNN and Web Audio API feature gaps - @TBD +- ❓ TBD +- βœ”οΈ Proposal: TBD + ## September 29, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200929T14) >πŸ”Ž **Scope:** Machine Learning Experiences on the Web: A **User’s** Perspective From 0c91fbbae29b4608d71fd39d99b4605408ee5ed7 Mon Sep 17 00:00:00 2001 From: Anssi Kostiainen Date: Fri, 18 Sep 2020 11:34:18 +0300 Subject: [PATCH 7/7] Address agenda TBDs in Sessions #2 and #3, shuffle training topics --- agenda/README.md | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/agenda/README.md b/agenda/README.md index 5d006f4..05a6fe9 100644 --- a/agenda/README.md +++ b/agenda/README.md @@ -48,14 +48,14 @@ ## Considerations for creating and deploying models -ℹ️ [Protecting ML models #67](https://github.com/w3c/machine-learning-workshop/issues/67) - @jasonmayes @tidoust @pyu10055 @jbingham -- ❓ Some ML providers need to ensure their ML models cannot be extracted from a browser app. -- βœ”οΈ Proposal: Investigate existing access control mechanisms for video, learnings from 3D assets. - ℹ️ [ML model format #74](https://github.com/w3c/machine-learning-workshop/issues/74) - @cynthia @jbingham @wchao1115 - ❓ There is no standard format for packaging and shipping ML models, model formats evolve rapidly. - βœ”οΈ Proposal: Initially focus on defining a Web API for accelerating established reusable ML operations instead of standardizing a model format. +ℹ️ [Protecting ML models #67](https://github.com/w3c/machine-learning-workshop/issues/67) - @jasonmayes @tidoust @pyu10055 @jbingham +- ❓ Some ML providers need to ensure their ML models cannot be extracted from a browser app. +- βœ”οΈ Proposal: Investigate existing access control mechanisms for video, learnings from 3D assets. + ℹ️ [In-browser training #82](https://github.com/w3c/machine-learning-workshop/issues/82) - @irealva @hapticdata @cynthia - ❓ The current in-browser efforts are focused on inference rather than training. - βœ”οΈ Proposal: Understand successful real-world usages (e.g. Teachable Machine) and target transfer learning as the initial training use case for related browser API work. @@ -70,7 +70,7 @@ - ❓ Should libraries for browsers and/or Wasm execution environments be able to target WebNN and WASI-NN together? - βœ”οΈ Proposal: TBD -ℹ️ [Heterogeneous parallel computing for the web #92](https://github.com/w3c/machine-learning-workshop/issues/92) - @jeffhammond @Kangz @abrown +ℹ️ [Heterogeneous parallel computing for the web #92](https://github.com/w3c/machine-learning-workshop/issues/92) - @jeffhammond @Kangz - ❓ How do the heterogeneous parallel computing abstractions fit in with the web platform? - βœ”οΈ Proposal: TBD @@ -94,12 +94,12 @@ ## Improving web developer ergonomics -ℹ️ [JS Operator overloading for Machine Learning #73](https://github.com/w3c/machine-learning-workshop/issues/73) - @TBD +ℹ️ [JS Operator overloading for Machine Learning #73](https://github.com/w3c/machine-learning-workshop/issues/73) - @cynthia @huningxin - ❓ Limitations in ECMAScript expressiveness impose ergonomics limitations for JS APIs on the web platform e.g. in vector matrix or tensor operations. - βœ”οΈ Proposal: TBD ℹ️ [WebGL garbage collection #63](https://github.com/w3c/machine-learning-workshop/issues/63) - @jasonmayes @Kangz @wchao1115 @huningxin -- ❓ Garbage collection in the WebGL API affects multiple ML libraries. +- ❓ Garbage collection in the WebGL API affects multiple ML libraries through side effects. - βœ”οΈ Proposal: Identify any improvements in graphics APIs to alleviate the GC issue, ensure purpose-built APIs designed around computational graph abstraction (e.g. WebNN) optimize GC from library usage perspective. ℹ️ [Neural network-oriented graph database #102](https://github.com/w3c/machine-learning-workshop/issues/102) - @WenheLI @@ -109,11 +109,11 @@ ## Developing interactive web experiences with ML ℹ️ [Action-Response Cycle bottlenecks in interactive music apps #97](https://github.com/w3c/machine-learning-workshop/issues/97) - @teropa -- ❓ Action-Response Cycle in interactive (music) apps must execute within 20 ms. +- ❓ Action-Response Cycle in interactive (music) apps must execute within 20 ms. Today, web developers need to do some API gymnastics to meet the requirement. - βœ”οΈ Proposal: Investigate inference in AudioWorklet context and media integration e.g. fast streaming inputs from MediaStream. -ℹ️ Noise suppression with DSP+DNN, WebNN and Web Audio API feature gaps - @TBD -- ❓ TBD +ℹ️ [Noise suppression with DSP+DNN, WebNN and Web Audio API feature gaps #100](https://github.com/w3c/machine-learning-workshop/issues/100) - @jmvalin @teropa @huningxin +- ❓ What areas needs work on the web platform to ensure noise suppression models perform? The need for primitives like Basic Linear Algebra Subprograms, Web Audio API enhancements to allow better analysis of waveforms? - βœ”οΈ Proposal: TBD ## September 29, 2020, 2pm UTC [πŸ—“οΈ](https://www.timeanddate.com/worldclock/fixedtime.html?iso=20200929T14)