Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions Modeling

Question

Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions Modeling

miladfa7 opened this issue 5 years ago · 9 comments

Is this a complete implementation of the paper (Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions Modeling)?

Answer 1 · 2019-07-28T16:12:11.000Z

Not yet, but the ddpg actor itself is working. Tomorrow I will add regularization and state representation (I deleted it recently for simplicity purposes). The dataset will be available soon. The only thing left from ML20 are the rewards (ratings) and if there are any problems with copyright I will release a tree based solution to generate synthetic ratings. I also don't know about the user embeddings just yet. But I wanted to make it more than just 'code for paper' by adding graph and geometrical learning support as well as overhauling the dataset

Answer 2 · 2019-07-29T05:42:42.000Z

very good
thanks

Answer 3 · 2019-08-21T09:46:15.000Z

New article released, be sure to check it out: https://towardsdatascience.com/reinforcement-learning-ddpg-and-td3-for-news-recommendation-d3cddec26011

Answer 4 · 2019-09-02T09:43:15.000Z

hi I am pursuing my master degree in Artificial Intelligence at the University of Zanjan. I am interested in AI in particular Recommendation Systems, and I opted this phenomenal area of AI as a research topic for my master thesis. I have chosen "recommendation system based deep reinforcement learning " as topic my master thesis. Please if possible, share if you have any ideas in this area. Thank you

…

________________________________ From: Mike Watts <notifications@github.com> Sent: Wednesday, August 21, 2019 9:46 AM To: awarebayes/RecNN <RecNN@noreply.github.com> Cc: milad farzalizadeh <milad.fa7@hotmail.com>; Author <author@noreply.github.com> Subject: Re: [awarebayes/RecNN] Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions Modeling (#2) New article released, be sure to check it out: https://towardsdatascience.com/reinforcement-learning-ddpg-and-td3-for-news-recommendation-d3cddec26011 — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#2?email_source=notifications&email_token=AGESODDHU3JYD6J5OJ2U4X3QFUFGRA5CNFSM4IHMMHQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4ZDF7A#issuecomment-523383548>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AGESODBIDL4HAB3F3JOMTLLQFUFGRANCNFSM4IHMMHQQ>.

Answer 5 · 2019-09-03T07:09:24.000Z

Most of the recommender systems use lstm for state representation. What if you use finite item state representation with novel importance sampling algorithms, (i.e.) state is sampled from learning to rank (LambdaRank, RankNet) or some Bayesian based algorithm.
Lstm is not the best idea for state representation/1D convolution. Wavenets switched to TCNs a while ago: “The fall of RNN / LSTM” by Eugenio Culurciello https://link.medium.com/EoheS55VFZ
P.S. if you happen to implement any of these, please push a commit

Answer 6 · 2019-09-03T17:18:47.000Z

https://trello.com/b/wnor4IZf/recnn

Answer 7 · 2019-09-07T15:30:50.000Z

Hi Dear What is the roadmap I need to follow to learn deep reinforcement learning? What is the best course for learning "deep reinforcement learning"? Thank you for guiding me

…

________________________________ From: Mike Watts <notifications@github.com> Sent: Tuesday, September 3, 2019 5:18 PM To: awarebayes/RecNN <RecNN@noreply.github.com> Cc: milad farzalizadeh <milad.fa7@hotmail.com>; Author <author@noreply.github.com> Subject: Re: [awarebayes/RecNN] Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions Modeling (#2) https://trello.com/b/wnor4IZf/recnn — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#2?email_source=notifications&email_token=AGESODHGFLDQRI34BE3MXKDQH2L7RA5CNFSM4IHMMHQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5Y5SDQ#issuecomment-527554830>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AGESODHQM5PKJ2HCHX3RBJDQH2L7RANCNFSM4IHMMHQQ>.

Answer 8 · 2019-09-07T16:41:46.000Z

Hi there, sweetheart!
I got some cool mentoring in MIPT here in Russia on general Deep Learning in the junior year of high school, but I studied the RL by myself. TBH when I wrote my first article I was not very familiar with Reinforcement Learning, just wanted to do a cool school project and tell about something.

Check out Yandexes Practical RL: github.com/yandexdataschool/Practical_RL
They have materials linked in the lesson folder as well as interactive notebooks

Read Andrej Karpathy's blog: http://karpathy.github.io/

Jonathan Hui has some cool blogposts on general RL: https://medium.com/@jonathan_hui/rl-deep-reinforcement-learning-series-833319a95530

You can visit our website: ttps://www.dlschool.org/?lang=en
Or/And fork the repo: https://github.com/DLSchool/dlschool_english

Answer 9 · 2019-09-07T17:02:46.000Z

Thank you very much Great help❤️🙏

…

________________________________ From: Mike Watts <notifications@github.com> Sent: Saturday, September 7, 2019 4:41 PM To: awarebayes/RecNN <RecNN@noreply.github.com> Cc: milad farzalizadeh <milad.fa7@hotmail.com>; Author <author@noreply.github.com> Subject: Re: [awarebayes/RecNN] Deep Reinforcement Learning based Recommendation with Explicit User-Item Interactions Modeling (#2) Hi there, sweetheart! I got some cool mentoring in MIPT here in Russia on general Deep Learning in the junior year of high school, but I studied the RL by myself. TBH when I wrote my first article I was not very familiar with Reinforcement Learning, just wanted to do a cool school project and tell about something.

________________________________ Check out Yandexes Practical RL: github.com/yandexdataschool/Practical_RL They have materials linked in the lesson folder as well as interactive notebooks

________________________________ Read Andrej Karpathy's blog: http://karpathy.github.io/

________________________________ Jonathan Hui has some cool blogposts on general RL: https://medium.com/@jonathan_hui/rl-deep-reinforcement-learning-series-833319a95530

________________________________ You can visit our website: ttps://www.dlschool.org/?lang=en Or/And fork the repo: https://github.com/DLSchool/dlschool_english — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#2?email_source=notifications&email_token=AGESODDDABJUTHRQVTBTY6TQIPKUVA5CNFSM4IHMMHQ2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD6E4ZLQ#issuecomment-529124526>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AGESODEP56CQQCNFIB4ODK3QIPKUVANCNFSM4IHMMHQQ>.