Research talk: Challenges in multi-tenant graph representation learning for recommendation problems
87
Microsoft Research334 тыс
Следующее
25.01.22 – 1732:08Популярные
91 день – 1 5741:23:36
24.01.23 – 1 7243:34Опубликовано 25 января 2022, 1:37
Speaker: Arun Iyer, Principal Researcher, Microsoft Research India
Recent research has shown that representations learned from user-user and user-item graphs can be used to improve recommendation performance. In this research, the recommendation model is often trained with representation learning. In project DEEGO, we aim to learn representations for various entities from multiple entity interaction graphs that can be used in various downstream recommendation scenarios. The various entities involved evolve at different rates. Additionally, the downstream recommendation scenarios may be either jointly trained along with the representation or trained in a way that is decoupled from representation learning. The latter is likely to be the most common scenario. These different nuances are challenges that we need to overcome to build a practical system. In this talk, we present the work that has been done so far and the challenges we need to solve in the future.
Learn more about the 2021 Microsoft Research Summit: Aka.ms/researchsummit
Recent research has shown that representations learned from user-user and user-item graphs can be used to improve recommendation performance. In this research, the recommendation model is often trained with representation learning. In project DEEGO, we aim to learn representations for various entities from multiple entity interaction graphs that can be used in various downstream recommendation scenarios. The various entities involved evolve at different rates. Additionally, the downstream recommendation scenarios may be either jointly trained along with the representation or trained in a way that is decoupled from representation learning. The latter is likely to be the most common scenario. These different nuances are challenges that we need to overcome to build a practical system. In this talk, we present the work that has been done so far and the challenges we need to solve in the future.
Learn more about the 2021 Microsoft Research Summit: Aka.ms/researchsummit
Свежие видео
3 дня – 2691:00
4 дня – 117 43413:00
5 дней – 6 4861:00
6 дней – 88 1621:02
Случайные видео
39 дней – 1 1270:57
217 дней – 2241:05
16.02.23 – 60416:21
13.12.21 – 262 0369:33
10.04.20 – 139 92010:06
17.02.10 – 100 3049:28
2 дня – 198 8729:06
3 дня – 2373:53
4 дня – 2 7641:30
6 дней – 2 0410:41
9 дней – 297 05516:28
10 дней – 380 5110:34
10 дней – 3542:52
11 часов – 3770:55
1 день – 207 86010:36