Seminar on Internet Technologies (Summer 2018): Difference between revisions

Line 51: Line 51:
|-
|-
| '''Reinforcement Mechanism Design'''   
| '''Reinforcement Mechanism Design'''   
Mechanism design is a modeling and algorithmic framework  to  design  and  optimize  mechanisms  in  dynamic  industrial  environments  where  a  designer can  make  use  of  the  data  generated  in  the  process to automatically improve future design.  Reinforcement mechanism design is  rooted  in  game  theory  but  incorporates  recent AI  techniques  to  get  rid  of  nonrealistic modeling assumptions and to make automated optimization feasible.  The framework can be applied on many key application scenarios, such as Baidu and Taobao, two of the largest mobile app companies in China. For the Taobao case, the framework automatically designs mechanisms that allocate buyer impressions for the e-commerce website; for the Baidu case, the frame-work automatically designs dynamic reserve pricing schemes of advertisement auctions of the search engine. Experiments show that the solutions outperform the state-of-the-art alternatives and those currently deployed, under both scenarios.
| Mechanism design is a modeling and algorithmic framework  to  design  and  optimize  mechanisms  in  dynamic  industrial  environments  where  a  designer can  make  use  of  the  data  generated  in  the  process to automatically improve future design.  Reinforcement mechanism design is  rooted  in  game  theory  but  incorporates  recent AI  techniques  to  get  rid  of  nonrealistic modeling assumptions and to make automated optimization feasible.  The framework can be applied on many key application scenarios, such as Baidu and Taobao, two of the largest mobile app companies in China. For the Taobao case, the framework automatically designs mechanisms that allocate buyer impressions for the e-commerce website; for the Baidu case, the frame-work automatically designs dynamic reserve pricing schemes of advertisement auctions of the search engine. Experiments show that the solutions outperform the state-of-the-art alternatives and those currently deployed, under both scenarios.
| Basic knowledge of machine learning, deep learning and big data analysis. Familar with mechanism design.
| Basic knowledge of machine learning, deep learning and big data analysis. Familar with mechanism design.
|Bo Zhao (bo.zhao@gwdg.de)
|Bo Zhao (bo.zhao@gwdg.de)
87

edits