Microtask crowdsourcing, as an efficient andeconomical method for a requester to outsource tasks to online workers, isbecoming increasingly popular in many domains, especially collecting labels forlarge-scale datasets. In microtask crowdsourcing, a requester usually needs toaccomplish three steps: firstly, recruit as many as possible workers from themarket; then, assign tasks to the workers based on their performance; lastly,reward good workers and meanwhile punish bad workers. For these three steps,various mechanisms have been proposed. Under certain assumptions about workers'responses to the rewards, these mechanisms can theoretically ensure workers tofollow the strategies desired by the requester and thus maximize the revenue ofthe requester. However, these assumptions may be violated in practice, whichcauses the failure of these theoretically elegant mechanisms. Thereby, recentstudies move their focus to the learning-based mechanisms which learn workers'models in an online fashion rather than simply assumin g one. In this thesis,we propose three novel learning-based mechanisms, each for one step, to pushforward the studies in this direction.

Zehong Hu has recently received Ph.D. degree fromNanyang Technological University (NTU). Now, he works as an Algorithm Expert atAlibaba Group. 

His researchinterests are in the areas of game theory, e-commerce and reinforcementlearning. During his Ph.D. studies, he has published several papers in top-tierconferences, including AAAI, IJCAI, NIPS and AAMAS. He has also severed as theorganizer of UMAP 2018.