While much research has been devoted to algorithm improvement of the machine learning model for multimedia applications, relatively little research has focused on the acquisition of massive multimedia datasets with strict data demands for model training. In this paper, we propose a Quality-of-Information (QoI) aware incentive mechanism in multimedia crowdsensing, with the objective of promoting the growth of an initial training model. We begin with a reverse auction incentive model to maximize social welfare while meeting the requirements in quality, timeliness, correlation, and coverage. Then, we discuss how to achieve the optimal social welfare in the presence of an NP-hard winner determination problem. Lastly, we design an effective incentive mechanism to solve the auction problem, which is shown to be truthful, individually rational and computationally efficient. Our evaluation study is carried out using a real multimedia dataset. Extensive simulation results demonstrate that the proposed incentive mechanism produces close-to-optimal social welfare noticeably, while accompanied by accelerating the growth of the machine learning model with a high-QoI dataset.

中文翻译：

---

QoI-aware 激励多媒体群体感知启用的学习系统

虽然很多研究致力于多媒体应用的机器学习模型的算法改进，但相对较少的研究集中在对模型训练有严格数据要求的海量多媒体数据集的获取上。在本文中，我们提出了一种多媒体人群感知中的信息质量（QoI）感知激励机制，目的是促进初始训练模型的发展。我们从逆向拍卖激励模型开始，在满足质量、及时性、相关性和覆盖范围的要求的同时，最大限度地提高社会福利。然后，我们讨论如何在存在 NP-hard 赢家确定问题的情况下实现最优社会福利。最后，我们设计了一个有效的激励机制来解决拍卖问题，这是真实的，个人理性和计算效率。我们的评估研究是使用真实的多媒体数据集进行的。广泛的模拟结果表明，所提出的激励机制显着产生了接近最佳的社会福利，同时伴随着具有高 QoI 数据集的机器学习模型的加速增长。