Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

第三季6场 AI ML Club 活动纪要 - AI ML Club #345

Open
utterances-bot opened this issue Sep 19, 2019 · 8 comments
Open

第三季6场 AI ML Club 活动纪要 - AI ML Club #345

utterances-bot opened this issue Sep 19, 2019 · 8 comments
Labels

Comments

@utterances-bot
Copy link

@utterances-bot utterances-bot commented Sep 19, 2019

第三季6场 AI ML Club 活动纪要 - AI ML Club

« 本次沙龙通知: #325 «

https://ai-ml.club/events/seminar-meeting-minutes-3-6/

This comment has been minimized.

Copy link
Member

@huan huan commented Sep 19, 2019 — with utterances

补充一个心得

Euclidean 距离和 Cosine 距离从数学角度来讲基本等价,除了 cosine 距离中,限制了向量的模值必须等于1。“模值等于1”这个约束,对训练有力,所以实际应用中,arc face 使用了 cosine 距离,比 facenet 产生了一定优势。(感谢范弘炜的分享)

AMC Logo Update

AMC Logo 中,已经将 "AI ML Club" 字样,改为了我们的域名:"AI-ML.Club"

@huan

This comment has been minimized.

Copy link
Member

@huan huan commented Sep 19, 2019

主席 Manual 需要修订

  • 新人流程:活动中,新人由邀请人负责加入 “AMC Open Forum - AMCOF” 微信群
  • 实习会员流程:活动中,第一次完成分享的新人,将升级为实习会员。有其邀请人负责将其加入“AI ML Club - AMC”会员群。(如果邀请人不在,则由当期主席负责)
  • 实习会员转正流程:参加了三次活动的实习会员(含三次),将有资格转正。转正流程:发送个人 profile 页面的 Pull Request 更新 https://ai-ml.club/people/GITHUB_USERNAME/ 下。PR Merge 后正式成为 AMC 会员
  • 实习主席转正流程:将完成了第一次轮值主席工作的实习主席,加入 Github Team: chairs,并在 team 中授予 maintainer 权限,便于未来升级其他主席。
  • PR Approve & Merge: 主席应该对自己任期内的 PR 负责,并在自己完成主席职责(发布活动纪要)之前,完成合并。对于无法合并的 PR ,应进行关闭。
@hwfan

This comment has been minimized.

Copy link
Contributor

@hwfan hwfan commented Sep 20, 2019

补充一个心得

Euclidean 距离和 Cosine 距离从数学角度来讲基本等价,除了 cosine 距离中,限制了向量的模值必须等于1。“模值等于1”这个约束,对训练有力,所以实际应用中,arc face 使用了 cosine 距离,比 facenet 产生了一定优势。(感谢范宏炜的分享)

AMC Logo Update

AMC Logo 中,已经将 "AI ML Club" 字样,改为了我们的域名:"AI-ML.Club"

弘 not 宏 :-)
(为防之后的人也写错,还是决定回复更正一下了哈哈)

另补充:欧氏度量和余弦度量的区别在于模值,而这个区别的影响主要是来自训练中的一个经验规律,即在人脸数据训练中,特征的模值往往代表样本的质量。这是因为最后得到的特征经过了BN模块的归一化处理,使得更一般化的特征会往中间集聚,更特殊化的特征会往周边弥散(笔者在MegaFace训练数据库上,也得到了能够佐证中这个规律的实验结果)。因此需要进行模值归一化,让网络对于不同质量的样本作等同的重视。

@huan

This comment has been minimized.

Copy link
Member

@huan huan commented Sep 20, 2019

- 范宏炜
+ 范弘炜

收到,fixed!

@huan huan added the blog comment label Sep 20, 2019
@huan

This comment has been minimized.

Copy link
Member

@huan huan commented Sep 20, 2019

@hwfan Please join our BUPT organization!

You've invited hwfan to Beijing University of Posts and Telecom! They'll be receiving an email shortly. They can also visit https://github.com/BUPT to accept the invitation.

@AlaskaaWong AlaskaaWong mentioned this issue Sep 20, 2019
3 of 7 tasks complete
@huan

This comment has been minimized.

Copy link
Member

@huan huan commented Sep 20, 2019

Please join our BUPT organization!

You've invited @AlaskaaWong to Beijing University of Posts and Telecom! They'll be receiving an email shortly. They can also visit https://github.com/BUPT to accept the invitation.

@newip

This comment has been minimized.

Copy link
Contributor

@newip newip commented Sep 22, 2019

Registered for the S3E7 event. Refer to @huan
BTW: Why can't I reply from ai-ml.club directly?

@huan

This comment has been minimized.

Copy link
Member

@huan huan commented Sep 22, 2019

@newip agree to refer to me.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
4 participants
You can’t perform that action at this time.