No abstract

Low engagement rates and high attrition rates have been formidable challenges to mobile apps and their long-term success. To date, little is known about how companies can scientifically detect user engagement stages and optimize corresponding personalized-targeting promotion strategies to improve business revenues. This paper proposes a new structural forward-looking hidden Markov model as combined with a randomized field experiment on app notification promotions. Our model can recover consumer latent engagement stages by accounting for both the time-varying nature of users’ engagement and their forward-looking consumption behavior. The structural estimates from the FHMM with the field-experimental data enable us to identify heterogeneity in the treatment effects. Additionally, we simulate and optimize the revenues of different personalized-targeting promotion strategies with the structural estimates. Personalized dynamic engagement-based targeting based on the FHMM can generate substantially higher revenues than the experience-based targeting strategy applied by current industry practices and targeting strategies based on alternative customer segmentation models. Overall, the novel feature of our paper is its proposal of a new personalized-targeting approach combining the FHMM with a field experiment to tackle the challenge of low engagement with mobile apps.

In this letter, we present a novel memristor-based restricted Boltzmann machine (RBM) system for training the brain-scale neural network applications. The proposed system delicately integrates the storage component of neuron outputs and the component of multiply-accumulate (MAC) in memory, allowed operating both of them in the same stage cycle and less memory access for the contrastive divergence (CD) training. Experimental results show that the proposed system delivers significantly 2770x speedup and less than 1% accuracy loss against the x86-CPU platform on RBM applications. On average, it achieves 2.3x faster performance and 2.1x better energy efficiency over recent state-of-the-art RBM training systems.

In this study, using a Bayesian learning model with a rich data set consisting of 2 million fine-grained GPS observations, we study the role of information observable by or made available to taxi drivers in enabling them to learn the distribution of demand for their services over space and time. We find significant differences between new and experienced drivers in both learning behavior and driving decisions. Drivers benefit significantly from their ability to learn from not only information directly observable in the local market but also aggregate information on demand flows across markets. Interestingly, our policy simulations indicate that information that is noisy at the individual level becomes valuable when aggregated across relevant spatial and temporal dimensions. Moreover, we find that the value of information does not increase monotonically with the scale and frequency of information sharing. Our results also provide important evidence that efficient information sharing can lead to a welfare increase because of potential market expansion. Efficient information sharing can bring additional income-generating opportunities that could be unfulfilled. Overall, this study not only explains driver decision-making behavior but also provides taxi companies with an implementable information-sharing strategy to improve overall market efficiency.

No abstract