Hong Kong Machine Learning Meetup Season 1 Episode 11

Clover Health will sponsor this meetup by providing the location, drinks and snacks! Thanks to them!
Clover Health (www.cloverhealth.com) will present on how they are improving the healthcare system using technology and machine learning.

Summary of the previous meetups:

• HKML S1E1 http://marti.ai/hkml/2018/07/18/hkml-s1e1.html
• HKML S1E2 http://marti.ai/hkml/2018/08/21/hkml-s1e2.html
• HKML S1E3 http://marti.ai/hkml/2018/10/09/hkml-s1e3.html
• HKML S1E4 http://marti.ai/hkml/2018/11/21/hkml-s1e4.html
• HKML S1E5 https://gmarti.gitlab.io/hkml/2018/12/19/hkml-s1e5.html
• HKML S1E6 https://gmarti.gitlab.io/hkml/2019/01/23/hkml-s1e6.html
• HKML S1E7 https://gmarti.gitlab.io/hkml/2019/02/20/hkml-s1e7.html
• HKML S1E8 http://gautier.marti.ai/hkml/2019/03/12/hkml-s1e8.html
• HKML S1E9 http://marti.ai/hkml/2019/04/17/hkml-s1e9.html
• HKML S1E10 http://marti.ai/hkml/2019/05/15/hkml-s1e10.html

We now have a LinkedIn Page: https://www.linkedin.com/company/hong-kong-machine-learning/ where you can share news or relevant information to the community.

Talks start at 7pm.

Programme:

TBA. Looking for more interesting speakers!

David Zhu - On Clover machine learning & data research
To be precised later.

Pranav will be introducing some key concepts for NLG (Natural Language Generation) and then show some of the novel methods that he developed for paraphrasing.

Firstly, he will give you an intro to seq2seq and subword models. After that, he will talk about paraphrasing through backtranslation. However, backtranslation is not good enough to generate diverse paraphrases. Hence, he will introduce concepts like "cold start" with a pretrained language model and diverse beam search.

He will then introduce his fun but peculiar technique on paraphrasing using graphical models using wavelets (think, Fourier Series). This is then plugged onto Graph2Seq to generate paraphrases with coherent order. Lastly, he will give a quick code walkthrough on recent literature and codebases, so you can do cutting-edge paraphrasing by yourself.

Pranav is currently the Machine Learning Head at Dayta AI and also splits his time for his side projects and research work. He has also published his research in top tier NLP conferences like ACL 2018 and NAACL 2019.

Benoit Goujon - How to help metro operators optimize the passenger experience using machine learning